A report, a proposal, a slide deck — you know you "have to make it," yet your hands freeze in front of a blank screen. The most time-consuming part of document creation isn't actually the writing itself; it's that first step of deciding "what to write, and in what order." In this chapter, you'll build the skill of having AI break that fear of the blank page, then speeding up structuring, drafting, turning it into slides, and finishing — all in one go. The key-point notes you made in the previous chapter, "Meetings, minutes, and summaries," become the raw material for your next document.

What this chapter will let you do

The goal: cut blank-page time to zero and focus on decisions

A skeleton in an instant
Have AI produce the outline for a plan, report, or proposal, and let the human handle what to keep or cut.
Drafts and slide conversion
You'll be able to ask for a first draft of each section, and conversion into a slide structure.
Boost persuasiveness
You'll have review lenses — conclusion first, numbers, concrete examples — for finishing a document that lands.

Beat the "fear of the blank page" with AI

Document creation is painful because it's the work of shaping the material in your head into a form that "gets the point across." Here AI's role is clear. Don't have AI make the finished product; have it produce a first draft, then edit on top of it as the human. Writing from zero versus editing something that's 70% done are worlds apart in psychological weight. First, hold this division of labor in mind: "the first draft to AI, the judgment to you."

😓 Making it from a blank page

You agonize over the structure, the opening, the phrasing. Getting started is the heaviest part, and you can't move until the deadline looms.

🚀 Editing from a first draft

AI presents the structure and a draft. The human just edits: "this is wrong," "add this." You can get moving right away.

Have it build the structure and outline

The iron rule of document creation is to start from the structure (outline). Start writing the body straight away and the story gets lost partway through. When you have AI produce a skeleton, the trick is to convey ① the type of document ② the purpose ③ the reader ④ the material you want included ⑤ a target length or page count. This alone sharply lowers the odds of getting a skeleton that misses the mark.

Example prompt — produce a proposal skeleton

Create an outline for an internal proposal for a new service.
・Purpose: get budget approval for next fiscal year
・Reader: executives (they value numbers and conclusions)
・Elements to include: the problem, the solution, expected impact, budget needed, schedule
・About 4–5 A4 pages
First, propose just the headings (chapter structure) as bullet points. For each heading, add a one-line note on what it covers.

The key is to have it produce just the headings first. Look at the big picture and adjust on the spot — "this chapter isn't needed," "swap the order" — before moving to the body, and you avoid major rework.

① Purpose in one sentence

"Get approval," "reach agreement," "build understanding" — hand over the document's goal first. If this wavers, everything wavers.

② Specify the reader

Executives, the front line, or an outside party? The emphasis and the wording change with the reader. Always add "who will see this."

③ Hand over material

Paste in your notes and numbers. The more material, the more concrete the content, and you avoid bland generalities.

💡 Have it produce several skeleton options. Ask "give me three versions with different angles," and structures you'd never have thought of get mixed in. Pick the best and combine the good parts, and you lock in a high-quality skeleton in a short time.

Have it write a first draft of the body

Once the structure is set, next comes the first draft of each section. Here too, rather than writing everything at once, writing section by section improves quality. Narrow the scope with "write the content of this heading in about 100 words," and the story won't sprawl thin, and edits get easier.

What's especially useful in drafting is converting back and forth between bullets and prose. Your thoughts are usually organized as bullets. Hand those over to have them turned into prose, and conversely have long prose compressed into bullets. Just this back-and-forth greatly changes how readable a document is.

📝 Bullets → prose

Hand over note fragments and ask "connect these into a polished explanation." The organization in your head becomes the manuscript.

📋 Prose → bullets

Hand over a long explanation and ask "into 3–5 bullet points." It instantly becomes material for slides or a summary document.

Example prompt — from notes to body text

Turn the following notes into the body of the "current problems" section of a report.
・Inquiry handling takes too long
・Response quality varies from person to person
・Past response history is hard to find
The reader is internal management. State the facts plainly, in about 150 words. Avoid exaggerated expressions.

Specify length and tone like "about 150 words" and "avoid exaggeration," and you get a draft that barely needs editing. Always read what comes back and fix anything factually off or overstated by your own hand.

⚠️ A draft is a draft. AI can invent plausible-looking numbers and examples on its own. Never use numbers, proper nouns, or quotes as-is — always verify them. The previous chapter's mindset, "a human always checks at the end," applies exactly the same to documents.

Reshape it into slide form

The report prose is done. But converting it into slides is another chore — and this is another conversion AI is good at. The key is not to paste the prose in as-is, but to have it broken down into "one slide = one message." Have each slide produced as a three-part set — "title," "key points (bullets)," and "what to say (spoken notes)" — and you get a form you can pour straight into a slide tool.

Example prompt — prose into a slide structure

Restructure the following report into slides for a 10-minute spoken presentation.
・Assume 7–8 slides total
・Each slide in the format "Title / Key points (up to 3 bullets) / What to say (2–3 sentences)"
・Put the conclusion at the start and next actions at the end
[paste the report body here]

This output becomes a blueprint for each individual slide. Keeping "what to say" separate also prevents the mistake of cramming too much text onto the slide.

Take the finished structure into the slide tool you're used to. You don't need to be bound by any one tool's steps. Just remember: lock in the "content and flow" with AI, and shape the look with the tool — this division is the fastest. Many slide tools have a feature that auto-builds slide frames from an outline, and pasting in the headings and bullets AI produced assembles the foundation in an instant.

STEP 1
Lock in the content with AI

Prepare the "title / key points / what to say" three-part set for every slide.

STEP 2
Pour it into the tool

Paste into the outline-import feature to auto-generate the slide frames.

STEP 3
Shape the look

Adjust colors, figures, and photos in the tool. The content is already done.

🖥 If you want the steps in more detail, how to create presentation slides with AI explains the concrete flow and the tools you can use.

Turn charts and numbers into words

Documents inevitably include tables and graphs, but just pasting numbers leaves the reader thinking "so, what's your point?" Here, have AI turn the key points of a table or graph into an explanation. Conversely, hand over a rambling explanation and ask it to "put this into a table," and it organizes it into a form you can compare at a glance. This two-way conversion raises a document's clarity across the board.

📊 Numbers → explanation

Paste a table or graph and ask "give me three key takeaways from these numbers, in prose." You can add words that put the graph's meaning into language.

🧮 Explanation → table

Hand over a comparison explanation and ask "into a table of items × options." Scattered information gets organized into a list that's easy to compare.

💡 Ask for "the one thing you most want to say." Ask "what's the single point this graph should emphasize?" and the axis you should convey comes into focus. Aggregating and analyzing the numbers themselves is explored in depth in the next chapter, "Data, spreadsheets, and analysis."

The finishing touches that boost persuasiveness

Once the content is in place, the last stage is polishing it into "a document that lands." Persuasive documents share a common form. Have AI review with that form in mind, and you can flush out weaknesses you'd struggle to notice yourself. In particular, these three are worth having AI check directly.

① Conclusion first

Busy readers want the conclusion first. Have it check that the structure puts the key point at the top of each section.

② Back it with numbers

"Cut by 30%" beats "greatly improved." Have it point out where you can add concrete numbers to abstract claims.

③ Add concrete examples

Generalities alone don't move people. Have it flush out where "for example" concrete cases are missing.

When you ask for a review, the trick is not just "take a look" but to specify the lens. Hand over evaluation criteria — "The reader is executives. Point out whether the conclusion comes first, whether claims are backed by numbers, and whether jargon is used without explanation, on these three points" — and the more criteria you give, the sharper the feedback. Also, having it anticipate objections, "list three concerns executives are likely to have about this proposal," works well for plugging holes in advance.

✅ Polishing how you ask changes the outcome. For the same document, the quality of the draft differs sharply with how you give instructions. For prompt patterns, the practical prompt engineering guide helps, and for the mindset of streamlining work overall, how to boost work efficiency with AI is a good reference.

Chapter summary
  • Document creation is "the first draft to AI, the judgment to the human." Instead of writing from a blank page, switch to editing a 70%-done draft.
  • First have it make the outline (headings only), keep or cut, then move to the body. Handing over purpose, reader, and material is the key to accuracy.
  • Draft section by section. Converting bullets ⇄ prose and prose ⇄ table raises a document's clarity.
  • Have slides broken into the "title / key points / what to say" three-part set; lock in the content with AI and shape the look with the tool.
  • Finish with conclusion first, numbers, and concrete examples. Specify the lens and have AI review, and a human always checks the numbers and proper nouns.

Once the document's frame and body are done, all that's left is the persuasive power of the numbers that support it. In the next chapter, Chapter 5, "Data, spreadsheets, and analysis," let's move on to consulting AI about spreadsheet aggregation, charting, and reading trends.