Turn a long slide deck into a useful pre-read
Thirty-two slides arrive the night before the meeting. A good pre-read tells people what they're being asked to decide. A bad one tells them the pilot is already approved and saving money.
What you'll make
a one-page pre-read that states the decision, summarises what’s proposed with slide numbers, keeps the conditions intact and lists what the deck doesn’t say.
What you'll need
the deck, who the pre-read is for and the decision if you know it, plus your work AI account. Upload the deck, or a PDF export of it, to ChatGPT, Claude or Gemini; or open it in PowerPoint and use Copilot on the ribbon. Check whether the tool read the speaker notes and any text inside images. About 15 minutes.
Example result.
Decision requested: approve a limited two-week pilot of a support tool (slide 3)
What's proposed: three support agents, dummy data only (slide 1);
nothing sent outside the team during the pilot (slide 2)
Cost: proposed cap of £200 (slide 2), not yet agreed
Not in the deck: any results, a vendor contract, who decides after the pilot
Questions for the meeting:
1. Who decides whether to extend after two weeks?
2. What would count as success?
3. Is the £200 cap inside the team's current budget?1. Tell AI what the reader needs to decide
Don’t start with “summarise these slides”. Say who will read the pre-read and what they have to decide. If the deck doesn’t say what the decision is, ask the AI to say so, rather than guess.
Write a one-page pre-read of this deck for [reader].
The decision is [decision, or "not stated: find it or tell me it's missing"].
Sections: decision requested, what's proposed, conditions and costs, not in the deck, questions for the meeting.
Give a slide number for every fact. Keep "proposed" and "approved" exactly as the deck uses them.
List any slides, notes or images you couldn't read.[reader] is who it’s for, such as “the head of customer service”; [decision] is what they’re being asked to agree.
Why this works: the decision gives the summary a job. Without it, you get a slide-by-slide recap, and the conditions that matter get the same weight as the agenda slide.
What you should see: five short sections with slide numbers, and a note of anything unread.
2. Check the headline facts against the slides
Open the deck beside the pre-read. For each fact, go to the slide it cites and check that the conditions came with it: “proposed”, “up to”, “dummy data only”, “subject to”. Then read the “not in the deck” section and ask whether anything is missing from it, such as results, approvals or contract terms. If the deck proposes improvements, keep them as proposals, not outcomes.
For each fact in the pre-read, quote the line from the slide it cites.
Mark any fact where the pre-read drops a condition or makes a proposal sound decided.
Add to "not in the deck" anything a reader would expect but the slides don't say.Why this works: quoting the slide line makes a dropped “proposed” obvious, and the second request catches the gap the first draft didn’t mention.
What you should see: each fact beside its slide line, with any dropped condition marked.
3. Fix any line that makes the proposal sound approved
Correct the marked lines, then read the pre-read once as the person deciding. Could they tell from this page alone what’s being asked, what it costs and what’s still unknown? If you asked for a Word or PDF version, open that file and check it there, not just in the chat.
Why this works: the reader of a pre-read hasn’t seen the deck. The page has to stand on its own.
What you should see: a page someone could decide from, with “proposed” still in place.
Check the result
- The decision requested is the first line, in the deck’s own terms.
- Every fact has a slide number, and the slide says it.
- “Proposed”, “pilot”, “dummy data” and other conditions survived the shortening.
- “Not in the deck” names results, approvals or terms that aren’t there.
The usual failure: shortening turns a proposal into a fact. “Proposed £200 cap” becomes “Budget: £200”, and “two-week pilot” becomes “the new tool”. Spot it by searching the pre-read for “proposed”: if the deck used it and the pre-read doesn’t, a condition has gone.
The pre-read turns [proposal] into [claim]. Slide [slide] says [slide wording].
Correct that line and the decision section, keep the slide reference,
and add anything this changes to "not in the deck".[proposal] is what the slide proposes; [claim] is how the pre-read put it; [slide] and [slide wording] are the slide number and its exact words.
Worked example: a support-tool pilot
Fictional example. A customer service team’s three-slide deck asks for a small pilot of a support tool. (It’s the practice deck in this guide’s task kit.)
Slide 1 Proposed pilot scope: two weeks · three support agents · dummy data only
Slide 2 Pilot boundaries: no external send · proposed cost cap £200
Slide 3 Decision requested: approve this limited pilot scope
Approval: not recorded · Measured results: not recordedFirst pre-read (wrong): “The team is rolling out the new support tool with three agents. Budget: £200.”
Repair: the repair prompt above for slide 2 (“proposed cost cap £200”), then for slide 3 (“approve this limited pilot scope”).
Repaired pre-read: the example result at the top of this page.
- “Rolling out” came from nowhere. The deck asks for a two-week pilot with dummy data; nothing in it is live, and slide 3 says approval isn’t recorded.
- “Budget: £200” read like money already agreed. “Proposed cap, not yet agreed” tells the reader it’s their decision.
- Slide 3 also says no results are recorded. The repaired pre-read puts that under “Not in the deck”, so nobody asks how the pilot went.
Task kit for this job: Download (ZIP) · Reusable instructions
Go deeper
Charts without readable values
If a slide has a chart but no numbers, ask for the data behind it or mark exact values as unknown. Don’t let the AI estimate precise figures from a picture of bars. See explain a chart without inventing.
A pre-read isn't a slide summary
A slide can introduce an idea with a picture and three words; a pre-read has to make sense without the presenter. Ask for full sentences with the context the reader needs, not a copy of every slide heading.
Keep a source trail for higher-stakes decisions
For a board or budget decision, keep each claim with its slide line, so anyone can check it later. The executive summary guide shows how to repair an unsupported headline without losing the short version.