Reports, decks and the structure problem

Coverage is not an argument

Ask for a deck on a topic and you get coverage: background, current state, challenges, opportunities, recommendations, next steps. Six headings that fit any subject on earth, evenly weighted, with no view in them. It looks like a deck. It does not do what a deck does, which is to move a room from one position to another.

This is not a limitation you prompt your way out of with better adjectives. The argument is the part that is yours. The tool is very good at everything that comes after it.

Work in the right order

1. Write the one sentence. What must the reader believe or do when this ends? We should close the Tuesday clinic and move the staff to Thursday. If you cannot write that sentence, no tool can, and producing slides first will hide the fact that you have not decided.

2. Write the spine, in your own words. Five to nine lines, each one a claim, in the order that makes the one sentence unavoidable. This is twenty minutes of thinking and it is the whole job. A useful test is that the spine should read as an argument on its own, out loud, with no slides at all.

3. Now bring the tool in. Give it the spine and the evidence. Ask it to challenge the order, name the weakest link, and say what a hostile reader will attack first. This is where it earns its place — not as a writer, as a rehearsal audience. Ask specifically: what will the finance director ask that this does not answer?

4. Then expand and format. One slide per spine line. Speaker notes from your evidence. A title on each slide that states the claim rather than naming the topic — "Tuesday attendance has halved since March", not "Attendance". Slide titles that assert rather than label are the single largest improvement available to most decks, and the tools do it well once you ask.

The same order for a report

Reports fail the same way and are rescued the same way. Structure first, in your words. Then the tool for expansion, transitions, an executive summary written after the body from the body, and the pass that most people skip: give it your finished draft and ask what a reader will not understand on first reading, and which paragraph is doing no work.

That last question is an unusually good use of the technology, because it is a text-shaped question about a text you supplied. The truth is in the room.

Format, practically

• Text to slides. Most tools will produce an outline you can paste straight into PowerPoint's outline view or Google Slides' import. Free alternatives that do this well: Marp turns plain markdown into slides, and Quarto does the same for reports and presentations with real citations.
• Free design. Canva's free tier covers most corporate deck needs. LibreOffice Impress opens and saves .pptx, so a colleague on Microsoft 365 sees what you sent. Google Slides is free with an account.
• Charts. Ask the code-running mode to make the chart from your actual file — real arithmetic, and it hands you back the code. Never let a chart be drawn from numbers that were typed out in a chat window.
• Images. Generated images in a work deck usually cheapen it. A plain chart, a screenshot, or a photograph of the thing you are describing carries more weight than a synthetic illustration, and it does not raise the question of what else was generated.

What to check before it leaves your hands

Decks are checked less carefully than letters, because they look finished and because nobody reads them in full. That is exactly why errors survive in them.

• Every number against its source. A figure in 40-point type is not more true than one in a paragraph.
• Every claim in a slide title. Assertive titles are stronger and they commit you.
• Every comparison. "Up 12%" against what, over what period, on what basis.
• The one sentence. Read the titles in order. If they do not add up to your one sentence, fix the spine, not the slides.

The uncomfortable question worth asking

Once a deck can be produced in eleven minutes, the constraint that used to limit how many decks existed is gone. That is not obviously a gain. Before you build one, it is worth asking whether this should have been three paragraphs in an email, or a five-line note, or a conversation.

The tools make production cheap. They do not make anybody's attention cheaper, and attention is the thing you were actually competing for.