Presentations and Exec Comms

Plan a presentation around one audience, one idea and one decision. Lead with the ask, show evidence with its uncertainty and rehearse difficult questions. Use AI for production and practice while retaining the recommendation and factual checks.

Why this matters

Funding, launches and renewals can turn on a short presentation. Weeks of analysis lose value when the recommendation arrives late, chart titles hide the point or uncertainty disappears from the claim.

Design the meeting around what the room must decide. Prepare a short version, evidence for likely objections and an honest follow-up when answers are missing. Chapter 34 covers writing, and Chapter 26 covers event keynotes and demos.

Planning: audience, objective, one idea

The audience is the hero

Nancy Duarte’s Resonate (2010) casts the audience as hero and presenter as mentor. Plan for the audience’s changed belief or action. Her slide:ology (2008) supplies the “glance test”: a slide should convey its point in roughly three seconds. Chapter 26 applies her current-versus-possible-world structure to keynotes.

Name the audience’s job: an operations team improving outcomes, an executive funding a portfolio or a seller preparing a deal. Define what they need to do and what blocks them. Updates focused only on your team often work better in writing.

Audience analysis

Most PMs skip audience analysis because they think they know the room. Do it in writing anyway. For each key person, answer:

• Role in the decision. Who decides, who advises, who can block, and who will carry the outcome into their own organization?
• What they know. Have they read the pre-read? Do they know the product’s metrics, or will “diagnosis acceptance rate” need a sentence of explanation?
• What they believe. What do they think is true today about this product, this market or you? Where does your message contradict it?
• What they fear. Margin, risk, reputation, headcount, a regulator, a customer escalation. The fear usually shapes the questions.
• What they need to leave with. A decision, a number they can repeat upward, a reassurance, a task.

Elizabeth Newton’s 1990 tapping study, popularized by the Heaths, shows how experts overestimate shared context. Test the opening on someone outside the team and ask them to repeat the decision you need.

The objective: know, feel, do

State what the audience should know, feel and do. “Inform leadership about Q3” is a topic. An objective names a change: understand inference-cost pressure, trust the recovery plan and approve the pricing pilot. Make the action observable.

The Big Idea

Duarte and Cole Nussbaumer Knaflic’s Storytelling with Data (2015) both recommend a one-sentence Big Idea before production. State your view and the audience’s stakes:

[Audience] should [act or believe] because [your point of view], and if they don’t, [what’s at stake].

A Big Idea needs an audience, position and consequence. “Beacon Agents update” lacks them. A scoped-launch recommendation tied to evidence and competitive delay gives the room a claim it can assess.

Knaflic adds the 3-minute story: if you had only three minutes, what would you say? It’s the Big Idea plus the two or three supporting points that matter most. Write it and time it aloud; it’s your plan for when the slot gets cut.

Choosing the mode before the deck

Choose the medium before building. Bezos’s 2017 letter advocates silently read narratives for decisions. Edward Tufte’s The Cognitive Style of PowerPoint criticizes fragmented reasoning and weak statistical displays, citing the Columbia investigation’s concerns about engineering slides. Dense prose and visual support serve different needs.

You don’t have to adopt either position whole. Choose by purpose:

Richard Mayer’s research on multimedia learning (summarized in Multimedia Learning, 2001) supports the middle rows: people learn better from words and pictures together than from words alone. It isn’t an argument for slides full of text that you then read aloud.

Storyboard before software

Storyboard full-sentence titles before using slide software or asking AI to build. Read only the titles to check the argument. Formatting should follow the reasoning.

Structuring for executives

Answer first

Minto’s Pyramid Principle groups evidence under supporting arguments and one governing answer. Its opening establishes the reader’s question:

• Situation: what the audience already knows and agrees with.
• Complication: what changed, or what makes the situation a problem.
• Question: the question the complication raises (usually “what should we do?”).
• Answer: your recommendation, which sits at the top of the pyramid, with grouped supporting arguments beneath it.

Use BLUF, bottom line up front: state the decision, recommendation and cost of delay first. Kabir Sehgal’s 2016 HBR article applies the military writing convention to business. Put context afterward.

Prepare for a five-minute slot. Duarte recommends findings, recommendation and action first, followed by discussion. Her “10% rule” suggests five summary slides for a fifty-slide appendix. Pull detail in response to questions, and answer a requested question first.

Pre-reads

For consequential tradeoffs, send a pre-read with the ask, context, options, recommendation, risks, reversal evidence and appendix or FAQ. Chapter 34 covers the writing.

Open by restating the ask and inviting questions from the reading. If advance reading is unreliable, reserve ten minutes for silent reading. Spend the remaining time on decisions.

Handling interruptions

Treat interruptions as information about a blocking concern or sufficient context. Stanford’s Matt Abrahams recommends agreeing a short opening slot, signposting the points, then paraphrasing questions, responding and resuming.

Answer in one sentence, then show the supporting slide or appendix. If the room already accepts the argument, move to the ask.

Data storytelling

Explore, then explain

Separate exploratory analysis from explanatory charts. For the audience, remove unnecessary detail, direct attention to the evidence and add an informative title, axis labels and source note. Knaflic emphasizes audience needs over chart decoration.

The single most useful habit is the action title. Every chart’s title is a full sentence stating the takeaway. “MTTR by month” names a metric. “MTTR fell after agent-assisted diagnosis rolled out” makes a claim, and the chart then has to support it. If you can’t write the action title, you don’t know why the chart is there.

One chart that proves it

For each Big Idea, find the one chart that proves it and design around it, with at most one context chart for the baseline. Highlight the data that carries the point, gray out the rest, and annotate the moment that matters (a launch, a pricing change, an incident) on the chart itself. Chapter 18 covers which product reports matter.

Findings are separate from recommendations

Separate findings from recommendations. State expected results before showing unwelcome evidence, understand stakeholders’ constraints and acknowledge that valid data can support several choices. Discuss alternatives where the action is contested.

Words of estimative probability

Sherman Kent’s 1964 CIA essay describes an estimate calling an attack a “serious possibility.” Authors and readers interpreted it across a wide probability range, despite agreeing on the words. Kent proposed linking verbal judgments to numeric ranges:

For a product deck, any word like “likely” or “strong chance” carries a number or range, or the slide states its scale. Executives will repeat your word, not your caveat, so pick the word you’d be comfortable hearing back in a board meeting.

Delivery and nonverbal communication

What Mehrabian actually found

Mehrabian’s 7% words, 38% voice and 55% facial-expression finding concerns communication of feelings and attitudes, especially conflicting signals. His Silent Messages (1981) does not claim that words carry only 7% of a presentation’s meaning.

When words and delivery conflict about a feeling, listeners may trust tone and expression. State confidence only where warranted and explain uncertainty specifically.

Power poses and nerves

A 2010 study by Dana Carney, Amy Cuddy and Andy Yap linked expansive “power poses” to hormones and behavior. Carney rejected those effects in 2016, citing small samples, flexible analysis and failed replications. Cuddy and colleagues defend narrower subjective effects, which remain contested. Treat posing as neither a verified hormone nor performance intervention.

Alison Wood Brooks’s 2014 paper found that reframing anxiety as excitement improved feelings and performance across singing, speaking and math tasks. This chapter does not assess replication, so treat the inexpensive technique as advice with limited evidence.

Craft habits

Beyond the research, a set of habits is worth teaching as practitioner advice rather than science:

• Pause after the key sentence. Say the ask, then stop. Silence gives the decision maker room to respond.
• Slow down on numbers. State the number, the unit and the comparison, then pause before moving on.
• Look at the decision maker when you state the ask. Look at the questioner when you answer a question.
• Keep still while listening to pushback. Nodding along while someone attacks your plan reads as agreement.
• Don’t read your slides. The room reads faster than you speak.

Virtual and recorded presentations

Video calls

Jeremy Bailenson’s 2021 “Nonverbal overload” argues that video calls impose tiring nonverbal patterns. Keep segments short, vary the activity, put the ask in writing, hide self-view and permit cameras off. Name responders to resolve ambiguous silence. Test audio, camera, screen sharing and captions.

Recorded and async video

Async updates and recorded walkthroughs are routine for PMs: a weekly update to engineering, a recorded demo for a customer who missed the call. These norms are practitioner habits rather than findings, and they follow from how people watch recordings: skimming, sped up, often muted.

• Put the ask or the headline in the first 30 seconds.
• Keep it short. TED’s guidance to its speakers notes that strong talks can run 10 to 12 minutes and that many are only 6 to 8, under strict time limits and with coaching beforehand. Your product update has less to say than a TED talk.
• Add chapter markers and captions.
• Post a written summary under the video, with the ask, the key numbers and links to sources, so an executive can get the point without pressing play.
• Look into the camera lens, not at the faces on screen, when you state the ask, and frame yourself at eye level with light in front of you.
• Re-record a rambling first take. A recording is optional viewing, and a slow start is where people stop watching.

Q&A, pushback and executive presence

Prepare for the hardest questions

Prepare ten hard questions. For each, record the likely asker, a one-sentence answer, evidence, an appendix reference and a follow-up owner for gaps. Rehearse with a skeptic or a harness playing relevant roles.

This book suggests a simple answer pattern for pushback, as a heuristic rather than a sourced framework:

1. Acknowledge: restate the concern in the asker’s terms, so they know you heard it.
2. Answer: one sentence.
3. Anchor: one piece of evidence, or the appendix slide.
4. Advance: return to the ask, or state what would change your recommendation.

Saying “I don’t know”

On Stanford’s 2022 Think Fast, Talk Smart, Lloyd Minor recommends admitting uncertainty, following up and updating views openly. “I don’t know, and I’ll have it by Thursday” works when the answer arrives Thursday. Keep guesses labeled.

Executive presence

The Center for Talent Innovation’s 2012 research surveyed 268 senior executives and nearly 4,000 professionals. Respondents described executive presence through gravitas, communication and appearance. Senior executives weighted these at 67%, 28% and 5%, and attributed 26% of promotion readiness to presence. These are perceptions, not measured causal effects.

Presence judgments also reflected unequal expectations. People of color more often reported a stricter code than white respondents (56% versus 31%), and 81% found the guidance unclear. Focus coaching on substantive behavior and question cultural conformity disguised as performance advice.

For a PM, the substantive parts are concrete: know your numbers without looking, own your recommendation and its risks, stay specific when challenged, say what you don’t know and when you’ll know it, and change your mind in public when the evidence warrants it.

How to apply it

Use this method when a decision, a renewal or a launch depends on the outcome. For a routine meeting, steps 1, 3 and 9 are enough.

Plan (one to two weeks out)

1. Name the decision and the decision maker. If nobody can say what the meeting exists to decide, propose it.
2. Write the presentation plan: audience analysis, objective, Big Idea, mode, slot and ask.
3. Write the 3-minute story and the 5-minute cut, and time both aloud.
4. Gather the evidence through the harness, pulling reports, evals and incident reviews through connectors, with a source for every number.

Build (one week out)

5. Storyboard the titles as full-sentence action titles, and check that the titles alone tell the story.
6. Structure answer first: ask and recommendation, Situation, Complication, Question, three arguments, risks and options. Everything else goes to the appendix.
7. Design the one chart for each argument, with an action title, highlighting, annotation and its uncertainty (range, sample size, method).
8. Write the pre-read if the decision has real trade-offs, and send it at least a day ahead.
9. Audit every number. Compare each figure on every slide with its source, and put the source in the speaker notes.

Rehearse (two to three days out)

10. Prepare the Q&A sheet. Ten hardest questions, each with asker, one-sentence answer, evidence, appendix slide and follow-up owner.
11. Run the murder board. Rehearse with a colleague or the harness playing the hardest people in the room.
12. Rehearse the delivery: a timed full run, a 5-minute run, and a recorded run reviewed for pace, filler words and slide reading. Learn the first and last 60 seconds by heart.
13. Check the setup, including a recorded fallback for any live demo.

Deliver and follow up

14. Open with the ask, and say the summary comes first and discussion follows.
15. Let the room steer: answer interruptions in one sentence, jump to the slide, return to the ask.
16. Close on the decision: what was decided, who owns the next steps, by when.
17. Send the follow-up within a day, including the answers you promised.

The AI lens

Presenting AI-powered products honestly

AI products are right some of the time, so the honest description of their quality is a distribution. The temptation is to show the best number and the best demo, which earns a launch and loses trust at the first failure nobody was warned about. Chapter 31 covers how to measure AI quality; these are the rules for presenting it.

• Show the range and the sample size. Anthropic’s November 2024 research on model evaluation recommends reporting the standard error and a 95% confidence interval for eval scores, and comparing models with paired differences. The executive translation: never show a single accuracy number without its range and how many cases it came from, and never claim an improvement smaller than the noise.
• Show the denominator. “The agent resolved hundreds of incidents” means nothing without how many it attempted, and how many it got wrong.
• Show failures alongside wins. Curated good outputs mislead even when every example is real. Show representative failures, and what the product does when it’s wrong.
• Separate offline evals from production outcomes. An eval score and a customer’s MTTR are different evidence. Label which is which.
• Use a model card for executives. Margaret Mitchell and colleagues’ Model Cards for Model Reporting (2019) proposed documenting a model’s intended use, its performance across relevant groups and its evaluation procedures. One slide can do the same job for an exec audience: what the product is for, what it isn’t for, how it was tested, how it performs by segment and its known failure modes.
• Use estimative language with numbers. “The agent is usually right” needs Kent’s discipline: say the rate and the range.
• Don’t anthropomorphize. Avoid “the AI understands” or “the agent knows.” Say what the system does: retrieves, ranks, proposes, acts after approval.
• Label demos. Say when a demo is scripted or uses prepared data, and keep a recorded fallback. Chapter 26 covers demos in depth.

Doing the work with AI

AI tools can draft decks in templates. Claude added PowerPoint file creation in September 2025, warning about internet-access data risks. Claude for Microsoft 365, Copilot and Gemini also support template- or theme-based creation. Capabilities change quickly. Use approved tools and retain the planning and verification workflow.

Keep the Big Idea, editorial cuts and factual audit with the presenter:

1. The harness drafts a presentation plan from your brief and the evidence it pulls through connectors.
2. You write the Big Idea and the ask yourself, and you choose between options if you asked for several.
3. The harness drafts the storyline as action titles. You cut.
4. The harness builds slides in the corporate template.
5. You verify every number against its source; the harness can list mismatches, but you check them.
6. You rehearse with AI. Tools such as PowerPoint’s Speaker Coach report on pace, pitch, filler words, wordiness and reading slide text, and a harness can play personas for Q&A.
7. The harness drafts the follow-up from your notes, and you correct it before it goes out.

Test the draft for generic claims, audit invented or mis-scaled numbers, and keep confidential plans and data in approved tools.

The Beacon case: the Beacon Agents launch decision

The plan. You start with the presentation plan. The CEO decides; the CFO and the head of security can each block. Your audience analysis finds three fears: margin for the CFO, a repeat of the hallucinated root cause for security, and, for the CEO, losing the category while Northlight waits. Your Big Idea: The exec team should approve launching Beacon Agents at Northlight Live for allowlisted, low-risk actions only, because that’s where our evidence and guardrails are strong, and waiting a year cedes the category while the margin problem stays unsolved either way.

The structure. Slide one is the ask: approve a scoped launch, with broader autonomy deferred to the first half of next year. Then the Situation (customers grow and stay, with NRR at 118%), the Complication (margin pressure from inference cost and the hallucinated root cause) and the Question (launch now or wait). Three arguments follow, each with one chart: diagnosis acceptance rate, with its interval and sample size; the guardrail story (new evals and the approval gate since the hallucinated root cause); and the margin plan, tied to the move from per-environment to per-resolved-incident pricing (Chapter 15). The competitive analysis and eval detail go to the appendix. A four-page pre-read goes out two days ahead.

Presenting the failure. The hallucinated root cause gets its own slide in the main flow: what happened, the blast radius, why the guardrail missed it, what changed, and the acceptance rate with its confidence interval before and after the fix. The head of security already knows the story, and burying it in the appendix would cost you the room.

Rehearsal. You ask the harness, connected to the eval dashboard and the incident review, to play the CFO, the head of security and the Observe group PM, who wants Beacon built on Observe data. The CFO persona asks whether per-resolved-incident pricing recovers margin at current inference cost. You don’t know, so the Q&A sheet records the honest answer: the pricing pilot is designed to find out, with results at the next quarterly review. The number audit catches two figures that don’t match the eval report, one rounded wrongly and one from an earlier run.

The meeting. Your slot is cut to fifteen minutes. On slide two, the CEO interrupts: “Is this safe?” You answer in one sentence (“Only allowlisted low-risk actions qualify, every step needs human approval, and every action is logged”), jump to the guardrail slide, take two questions from security, and return to the ask. The decision is a conditional yes: launch with the allowlist, and come back with pricing-pilot results before any expansion of autonomy. Your follow-up goes out that afternoon.

The customer version. At a quarterly business review with Marcus, the VP of IT Operations, the hero is his organization. The Big Idea is about his team’s MTTR and on-call load, from his own data, shown as ranges with incident counts. The guardrail slide stays; the margin slide goes, because margin is Northlight’s problem.

Product 360 artifacts

# Presentation plan · [Meeting] · [Date]

**Presenter:** [PM]   **Slot:** [minutes]   **Format:** In person / Virtual / Hybrid / Recorded

## The decision
- Decision this meeting exists to make:
- Decision maker:
- People who can block it:
- Cost of not deciding (or of deciding late):

## Audience analysis
| Person | Role in the decision | What they know | What they believe today | What they fear | What they need to leave with |
|---|---|---|---|---|---|
| | | | | | |

## Objective
- They will **know**:
- They will **feel**:
- They will **do**:

## Big Idea (one sentence)
[Audience] should [act or believe] because [our point of view], and if they don't, [what's at stake].

## The hero
Who changes as a result of this presentation, and what stands in their way:

## Mode
Pre-read plus discussion / Talk with visuals / Written update only / Async video. Why this mode:

## The 3-minute story
[Big Idea plus the two or three most important supporting points, written as you'd say them]

## The 5-minute cut
What survives if the slot shrinks:

## Evidence needed
| Claim | Evidence | Source | Owner | Ready? |
|---|---|---|---|---|
| | | | | |

## The ask (exact words)
"[...]"

## Success signal
How we'll know afterward that the objective was met:

# Executive briefing · [Topic] · [Date]

**Presenter:** [PM]   **Decision maker:** [Name]   **Pre-read sent:** [Date]

## 1. The ask and recommendation (first 60 seconds)
- We are asking you to: [decision]
- We recommend: [option], because [one sentence].
- If we don't decide by [date]: [consequence].
- Format note to say aloud: "Summary first, then discussion. Detail is in the appendix."

## 2. Context (Situation, Complication, Question)
- Situation (what we all agree on):
- Complication (what changed):
- Question (what we must decide):

## 3. Supporting arguments (one action title and one chart each)
1. [Action title] · Evidence: [chart or data] · Source:
2. [Action title] · Evidence: · Source:
3. [Action title] · Evidence: · Source:

## 4. Options and trade-offs
| Option | What we get | What it costs | Main risk | Our view |
|---|---|---|---|---|
| Recommended | | | | |
| Alternative | | | | |
| Do nothing | | | | |

## 5. Risks and what would change our mind
- Risk · Mitigation · Owner
- We'd change our recommendation if: [evidence]

## 6. Next steps if approved
| Step | Owner | Date |
|---|---|---|
| | | |

## Appendix index
| Slide | Covers | Likely question it answers |
|---|---|---|
| A1 | | |

# Data story · [Topic] · [Audience] · [Date]

## Big Idea
[One sentence: point of view plus what's at stake]

## What the audience expects to see
[Their current belief, named before the new finding]

## The chart that proves it
- Action title (full-sentence takeaway):
- Chart type and why:
- What to highlight (color or emphasis):
- What to gray out or remove:
- Annotations (events, launches, changes):
- Source and date range:

## Context chart (optional, at most one)
- Action title:
- Why the main chart needs it:

## Uncertainty
- Range or confidence interval:
- Sample size (n) and time period:
- Method in one line:
- Estimative words used, with their numeric meaning:
- What this data can't tell us:

## Findings (what the data shows)
1.
2.

## Recommendation (what we propose to do; kept separate from findings)
-

## If the audience accepts the data but rejects the recommendation
Options to offer:

# Q&A prep · [Meeting] · [Date]

| # | Question (in the asker's words) | Likely asker | One-sentence answer | Evidence | Appendix slide | If I don't know: who follows up, by when |
|---|---|---|---|---|---|---|
| 1 | | | | | | |
| 2 | | | | | | |
| 3 | | | | | | |
| 4 | | | | | | |
| 5 | | | | | | |
| 6 | | | | | | |
| 7 | | | | | | |
| 8 | | | | | | |
| 9 | | | | | | |
| 10 | | | | | | |

## The question I most hope nobody asks
- Question:
- Honest answer:

## Pushback pattern
Acknowledge (restate the concern) · Answer (one sentence) · Anchor (one piece of evidence) · Advance (back to the ask, or what would change my view)

## Questions I won't answer in this room, and why
[For example, unannounced roadmap in a mixed audience. Prepared redirect:]

# [Product or feature]: how well it works · [Date]

## What it's for
[Intended use, in one sentence]

## What it isn't for
[Uses we don't support or recommend]

## How we tested it
- Eval set: [description, size n, how built]
- Production data: [period, number of cases]
- Offline eval and production results are labeled separately below.

## How it performs
| Measure | Result | 95% interval | n | Offline or production |
|---|---|---|---|---|
| | | | | |

## By segment
| Segment | Result | Interval | n |
|---|---|---|---|
| | | | |

## Known failure modes
| Failure | How often (with range) | What the product does when it happens | Fix status |
|---|---|---|---|
| | | | |

## Representative examples
Two successes and two failures, chosen to be typical rather than best.

## Wording rules on this slide
No single number without its range and n. No improvement claimed that is smaller than the interval. No "understands," "knows" or "thinks."

# Rehearsal checklist · [Meeting] · [Date]

## Content
- [ ] Full run, timed: [minutes] against a slot of [minutes]
- [ ] 5-minute cut run, timed
- [ ] 3-minute story said aloud without slides
- [ ] First 60 seconds and last 60 seconds said without notes
- [ ] Titles-only read: the action titles alone tell the story
- [ ] Someone outside the team says back the ask correctly

## Numbers
- [ ] Every number checked against its source; source in speaker notes
- [ ] Every chart's scale and axis checked
- [ ] Every estimative word ("likely," "most") has a number or range
- [ ] AI results show range and sample size

## Delivery
- [ ] Recorded run reviewed (or AI rehearsal report) for pace, filler words and reading slides
- [ ] Pause planned after the ask
- [ ] Numbers slowed down: number, unit, comparison

## Q&A
- [ ] Q&A prep sheet complete, ten questions
- [ ] Murder board run with the hardest personas (human or AI)
- [ ] Appendix slides findable in under ten seconds

## Setup
- [ ] Virtual: camera, audio, screen share, captions tested; self-view hidden
- [ ] Ask written in the chat or on screen for virtual audiences
- [ ] Demo: recorded fallback ready; scripted or prepared data labeled
- [ ] Pre-read sent [date]; follow-up template ready
- [ ] Deck built only in tools approved for this data

Prompts

Goal: Get to one sentence that takes a position, names the audience and states what's at stake.
Paste in: the presentation plan (or your notes on audience, decision and slot) and a summary of the evidence. If the evidence includes confidential financials or customer data, use a tool your company has approved for it.

I'm presenting to [audience] to get [decision]. Here is my evidence: [evidence summary].

1. Draft three candidate Big Ideas in the form "[Audience] should [act or believe] because [point of view], and if they don't, [what's at stake]." Make them genuinely different positions, not rewordings.
2. For each, list the evidence above that supports it and any evidence that cuts against it.
3. Now act as [the most skeptical decision maker, e.g. the CFO]. For each candidate, say in two sentences why it doesn't matter to you or why you don't believe it.
4. Tell me which candidate survives best, and what evidence I'd need to make it stronger.

Use the supplied evidence and identify gaps. State when it supports no strong position.

Check the output: you choose the Big Idea, not the model. Ask yourself whether a competitor's PM could have written the same sentence. If so, it's too generic.

Goal: Turn a deck written in the order you did the work into one that leads with the ask.
Paste in: the deck's slide titles and speaker notes (or the whole deck), the decision you need, and the slot length.

Here is a draft deck for a [length]-minute meeting with [audience]. The decision I need is [decision].

1. Rewrite the first slide so it states the ask, my recommendation and the cost of not deciding, in no more than 40 words.
2. Write a Situation, Complication, Question in one sentence each.
3. Group the remaining content into at most three supporting arguments, and give each one a full-sentence action title.
4. Move everything else to an appendix and list each appendix slide with the question it would answer.
5. Write the 5-minute version: which slides survive and what I'd say on each.
6. Flag any slide title the content on that slide doesn't support.

[deck]

Check the output: read only the new titles in order. If they don't tell the story and make the ask, the restructure failed. Make sure nothing in the appendix is something the decision maker needs before deciding.

Goal: Make every chart say its point and carry its uncertainty.
Paste in: each chart (image or underlying data), its current title, the source and the sample size.

For each chart below:
1. Write an action title: a full sentence stating the takeaway, supported by what the chart shows.
2. Tell me what to highlight and what to gray out or remove.
3. Suggest one annotation that would help the audience see the point.
4. State the uncertainty: the range or interval, n, and the time period. If I haven't provided enough to state it, tell me what's missing.
5. Replace any vague probability word (likely, probably, most) with a number or range, or flag it.
6. Flag any title where the chart doesn't support the claim, or where the scale could mislead.

[charts, sources, sample sizes]

Check the output: confirm each action title against the data yourself. Reject any title that claims more than the chart shows, or a change smaller than the stated range.

Goal: Hear the hardest questions before the meeting does.
Paste in: the pre-read or deck, and a short description of each key person (role, what they care about, what they fear). Use an approved tool if the material is confidential.

Act as each of these people in turn: [persona 1, e.g. "CFO who cares about gross margin and payback"], [persona 2, e.g. "head of security who approves anything touching production"], [persona 3, e.g. "peer PM who disagrees with our strategy"].

For each persona:
1. List the three questions you'd ask first, in your own words.
2. Name the one claim in this material you find least credible, and why.
3. Say what would make you say no.

Then step out of character. Rank the ten hardest questions across all personas, and for each draft a one-sentence answer using only the evidence in the material. Where the material has no answer, write "No answer in the material."

[material]

Check the output: the personas are a rehearsal aid, not a prediction. Ask a real colleague who knows the decision maker whether these are the right questions. Fill the "no answer" gaps with real evidence or an honest "I don't know, I'll follow up by [date]."

Goal: Catch invented, stale or mismatched numbers before the deck leaves your hands.
Paste in (or connect the harness to): the deck and every source report it draws on.

Compare every number, percentage, date and chart value in this deck with the source documents provided.

Produce a table with: slide number, the figure as shown, the matching figure in the source, the source name and location, and a status of Match, Mismatch, Rounded, Stale (from an older version of the source) or No source found.

Then list:
1. Every claim on a slide that has no source at all.
2. Every chart whose axis or scale could make a difference look larger or smaller than it is.
3. Every AI-quality number shown without a range or sample size.

Don't correct the deck yourself. Report only.

[deck]
[sources]

Check the output: spot-check at least three "Match" rows by hand, since the audit can be wrong too. Fix every Mismatch and No source found before sending.

Goal: Get specific delivery feedback from a recorded rehearsal or transcript.
Paste in: the transcript of a timed rehearsal (from your meeting tool or a rehearsal feature), the slot length and the intended ask.

Here is a transcript of my rehearsal for a [length]-minute presentation. The ask is [ask].

1. How many seconds in does the ask first appear? Quote the sentence.
2. List filler words and hedges ("just," "sort of," "I think maybe") with counts.
3. Find places where I explain a number without stating its unit, comparison or range.
4. Find any passage over 60 seconds with no clear point, and suggest a one-sentence replacement.
5. Find places where I seem to be reading slide text aloud.
6. Write my opening and closing 60 seconds as tighter versions, in my own words and style.

[transcript]

Check the output: re-record and compare. Keep your own phrasing where it's clear; the goal is a tighter you, not a script someone else would say.

Goal: Send a clear record of the decision and next steps within a day.
Paste in: your notes or the meeting transcript, the original ask, and the Q&A items you promised to follow up on. Use an approved tool for meeting recordings.

From these notes, draft a short follow-up email to the attendees:
1. The decision made (or not made), in one sentence, in the decision maker's terms.
2. Conditions attached to the decision.
3. Next steps as a table: step, owner, date.
4. Answers to any questions I promised to follow up on; where I don't have an answer yet, say when I will.
5. Anything I said in the meeting that I now need to correct.

Keep it under 200 words. Don't add commitments that weren't made in the meeting.

[notes or transcript]

Check the output: confirm the decision wording with the decision maker if there's any ambiguity. Check that every owner named actually agreed to the step.

Anti-patterns

Checklist

• The decision and decision maker are named, and a presentation plan has audience analysis and a know, feel and do objective.
• The Big Idea is one sentence with a position, an audience and a stake, and the mode was chosen on purpose.
• The 3-minute story and the 5-minute cut are written and rehearsed.
• The first slide or paragraph states the ask, the recommendation and the cost of not deciding.
• Every chart has a full-sentence action title that its data supports.
• Every number has a source in the speaker notes, and the deck has been audited against sources.
• Probability words carry numbers or ranges; AI results show range, sample size and failure modes.
• Findings are separated from recommendations, and options are ready if the room accepts the data but rejects the action.
• The Q&A prep sheet covers ten hard questions, including the one you most hope nobody asks.
• A murder board (human or AI) has been run against the deck.
• Virtual setup, demo fallback and captions are tested.
• Confidential material went only into approved AI tools.
• A follow-up with decisions, owners and dates goes out within a day.