Client Presentation Prompt for Consultants

What this prompt does

A client presentation prompt turns finished work — an analysis, a design, a market read — into the slide-by-slide structure for the meeting where the client decides what happens next. Its input is a recommendation you already believe; its output is the order in which a specific room hears it.

That's a different job from winning the engagement, and the two decks fail differently. A pitch fails when the client doesn't buy you. A delivery presentation fails when the client sits through forty minutes, says "great work, lots to think about," and decides nothing — usually because the deck was a tour of everything you did rather than an argument for what they should do.

"We did a lot of work" is the message most delivery decks actually send. The fix is structural, and it's old. Chris Anderson's advice on structuring a talk is built around a single throughline the audience can follow, and Michael Alley's assertion-evidence method makes every slide headline a full-sentence claim with the slide as its evidence — "Churn is concentrated in month two," never "Findings." The prompt below enforces both, plus the rule consultants resist most: the recommendation goes at the front.

The prompt

Build a slide-by-slide presentation outline from the inputs below. Follow every rule.

**The engagement in one sentence:** [WHAT YOU WERE HIRED TO DO]
**Your recommendation in one sentence:** [THE ANSWER. If you can't write this line yet, stop — finish the thinking before the deck.]
**The decision you need from the room:** [THE SPECIFIC THING — approve X, choose between A and B, sign off on the budget]
**Who's in the room:** [NAMES AND ROLES — who decides, who influences the decider, who can veto]
**What they already believe:** [THE EXPECTATION you're confirming or overturning]
**Time slot:** [MINUTES, and how much of it should be discussion]
**The evidence:** [YOUR 3–5 STRONGEST FINDINGS, with the real numbers]
**The two questions you're dreading:** [BE HONEST — each one gets an appendix slide]

Return a numbered slide list. For each slide give:
- A headline written as a full-sentence claim, never a label. "Support tickets doubled after the pricing change" — not "Findings".
- What's on the slide, in one or two lines.
- What you say out loud that isn't on the slide, in one line.

Rules:
- The recommendation appears in full no later than slide 3. The deck is not a mystery novel.
- At most one slide per two minutes of talk time. Cut until it fits.
- Methodology, process, and the story of how hard you worked go in the appendix. They come out only if someone asks.
- Every number must come from the evidence input. Invent nothing.
- One message per slide. If a headline needs an "and", it's two slides.
- The final content slide restates the ask as the specific decision, with what happens in the first two weeks after a yes.
- After the main deck, add one appendix slide per dreaded question, built to answer it in under a minute.

How to use it

The recommendation line does most of the work. If you can't state it in one sentence, you don't have a deck problem yet — you have a thinking problem, and no slide order will hide it.

Fill in "who's in the room" with actual names and what each person cares about, because the same recommendation gets a different slide 4 depending on whether the skeptic in the room is the CFO or the head of product. The model can't know that. You do.

Be honest in the dreaded-questions field. It's the highest-value input on the list. A question you've built an appendix slide for is a question you want someone to ask; the same question with no slide behind it is the moment the meeting gets away from you.

And keep the deck and the leave-behind separate. A deck dense enough to read without you is too dense to present. If the client needs a document afterwards, write one — that's a different artifact with a different prompt.

What each part of the deck has to do

Example output (partial)

Inputs said: pricing review for a 40-person B2B SaaS company, recommendation to raise the entry tier from $29 to $39 and grandfather existing customers for 12 months, decision needed is board-level sign-off, 30-minute slot, CFO expects the opposite (she suspected the entry tier was priced too high).

Slide 2 — "Raise the entry tier to $39; grandfather current customers for 12 months." On the slide: the recommendation and the two numbers that anchor it — trial-to-paid conversion held at 8.1% through the last two price tests, and 71% of churned entry-tier customers cited a missing feature, not price. Say out loud: "This is the opposite of what we expected going in, and the next four slides are why we changed our minds."

Slide 3 — "Price was named by 11% of churned customers; the missing-integrations complaint was named by 71%." On the slide: one chart, churn reasons from the 214 exit surveys. Say out loud: where the survey data came from and its one real limitation — it undercounts silent churners.

Appendix A — "What if competitors undercut us at $39?" Built for the CFO's likeliest push-back: the three nearest competitors' entry prices ($35, $42, $49) and what happened to conversion during the March test.

The model put the reversal of the CFO's expectation on slide 2, in the spoken line rather than on the slide — named early, so it reads as confidence instead of defensiveness.

Variations

For presenting design work

Designers lose the room in the reveal. Add a rule: the problem being solved gets a slide before any artwork appears, and each design slide's headline states what the design does, not what it looks like. Rationale before reveal — otherwise the client critiques fonts while the strategy goes unexamined.

For a listing presentation

A real estate listing presentation is a hybrid: it has to sell you, but it lives or dies on one recommendation — the price. Run this prompt with the pricing recommendation as the throughline and the comparable sales as evidence, then rehearse the spoken side with the sales script prompt. The homeowner's dreaded question is nearly always "the agent yesterday said they'd list it higher," and it deserves the best appendix slide you've got.

For a deck that gets sent, not presented

Async changes the rules: there's no "say out loud" column, so ask for headlines that carry the full argument on their own and a one-slide summary up front. It drifts toward being a document — past a certain point, write the document instead.

For an internal steering meeting

Swap the client for the sponsoring executive and add an input for what's changed since the last update. Steering decks rot into status theater without it.

Common pitfalls

• Don't structure the deck as the story of the engagement. Chronology is the order the work happened in, which is almost never the order the argument should arrive in.
• Don't accept label headlines. "Methodology," "Findings," "Next steps" — a slide title that could sit in any deck is doing no work in yours.
• Don't end on "Questions?". End on the decision, then take questions with the appendix loaded.
• Try instead: read just the headlines of the finished outline, top to bottom. They should work as a complete argument on their own — that's the assertion-evidence test, and it takes thirty seconds.

Who uses this prompt

• Consultants: the deck at the end of the engagement is the engagement, as far as the client remembers it. More in prompts for consultants.
• Designers: presenting work to clients who judge the rationale by the slide order it arrives in. See prompts for designers.
• Real estate agents: listing presentations where one number is the whole recommendation. See prompts for real estate agents.

Both ChatGPT and Claude handle this structure well, and Anthropic's prompt engineering guidance explains the pattern that makes it work: the rules ban the easy outputs — label headlines, chronological order, a buried ask — so the model has to commit to an argument. The slide-3 rule gets the most resistance from people trying it, and it's the one clients thank you for. Nobody has ever complained that a consultant got to the point too quickly.