Why “Make Me a Presentation” Fails

Type “make me a presentation about our Q4 results” into any AI presentation tool, and you’ll get… something. Probably 10 slides. Probably with charts. Probably using the word “leverage” somewhere embarrassing.

The problem isn’t the AI. The problem is the prompt. “Presentation” is a container format, not a genre. A pitch deck has different DNA than a training module. A quarterly business review has different DNA than a conference keynote. Asking for “a presentation” is like asking a chef for “food” — technically you’ll get fed, but it won’t be what you wanted.

Scenario prompting is the skill of specifying context so precisely that the AI output aligns with your actual needs — audience, purpose, format, tone, and structure all baked into the request. Here’s how to do it for every major presentation scenario.

The Anatomy of a Good Scenario Prompt

Every effective presentation prompt should address these five dimensions:

  1. Role: Who is the presenter? (CEO, sales lead, product manager, teacher, consultant)
  2. Audience: Who is receiving this? (investors, executives, customers, students, internal team)
  3. Goal: What should happen after the presentation? (funding decision, purchase, understanding, approval)
  4. Format: What kind of presentation is this? (pitch deck, report, training, keynote)
  5. Constraints: What are the non-negotiables? (slide count, time limit, brand rules, must-include content)

A prompt that hits all five dimensions is 10× more likely to produce useful output than a vague topic description. Let’s see this in action across different scenarios.

Scenario 1: The Investor Pitch Deck

The highest-stakes presentation most people will ever create. Investors see hundreds of decks. Yours needs to cut through.

Weak prompt: “Make me a pitch deck for a food delivery startup.”

Strong prompt: “You are a founder pitching to seed-stage venture investors. The audience has 15 minutes and sees 20+ decks per week. Goal: secure a follow-up meeting. Create a 10-slide pitch deck for a B2B food delivery logistics platform. Include: problem slide with a specific market statistic, solution slide showing our tech differentiator (real-time route optimization), market size (TAM/SAM/SOM), business model, competitive landscape with a 2×2 matrix, traction with actual numbers (revenue, customers, growth rate), team slide highlighting founder expertise, and a clear ask slide. Tone: confident but not arrogant. Data-heavy but not dense. Use specific numbers even if placeholders — I’ll replace them. Avoid buzzwords like ‘revolutionary,’ ‘disruptive,’ or ‘game-changing.’”

Why this works: It specifies slide structure (not just topic), defines tone constraints, mandates data specificity, and acknowledges the investor context (15 minutes, competing with 20 other decks). The AI now has guardrails that prevent the generic pitch-deck-template output.

Key prompt elements for pitch decks:

  • Always specify “ask slide” — what you want from the audience
  • Demand specific numbers (even placeholders) — vague claims kill credibility
  • Mention competitive format (comparison table, 2×2 matrix, feature checklist)
  • Set tone guardrails — pitch tone gone wrong is the #1 AI failure mode

Scenario 2: The Quarterly Business Review (QBR)

Internal QBRs have a specific rhythm: context, performance, analysis, forward look. The audience already knows the business — they want insight, not introduction.

Weak prompt: “Create a Q4 business review presentation.”

Strong prompt: “You are a VP of Sales presenting Q4 results to the executive leadership team. The audience cares about: did we hit targets, why or why not, and what changes next quarter. Create an 8-slide deck. Structure: Q4 targets vs. actuals (traffic-light indicators: green/red), revenue breakdown by region (bar chart), top 3 wins with root causes (not just ‘we did well’ — ‘we did well because X’), top 3 misses with corrective actions, pipeline health for Q1, and one strategic recommendation requiring leadership approval. Tone: direct and analytical. No corporate filler. If we missed a target, say ‘we missed’ — don’t say ‘we experienced headwinds.’ Include data placeholders in [brackets] for me to fill.”

Why this works: It speaks the language of internal reporting (traffic lights, root causes, pipeline health) and sets a tone that matches executive expectations (direct, analytical, no sugar-coating).

Key prompt elements for QBRs/reports:

  • Specify the decision the report should drive
  • Demand “why” analysis, not just “what” reporting
  • Set candor expectations — internal audiences hate spin
  • Use internal terminology your company actually uses

Scenario 3: The Sales Demo or Proposal

Sales presentations need to bridge “here’s what we do” and “here’s why you should buy.” The worst AI-generated sales decks are all about the vendor and not about the prospect.

Weak prompt: “Create a sales presentation for our CRM software.”

Strong prompt: “You are an account executive presenting to the VP of Sales at a mid-market manufacturing company. They’re currently using spreadsheets for pipeline management. Goal: move them from ‘interested’ to ‘trial commitment.’ Create a 12-slide deck. Structure: acknowledge their current pain (pipeline visibility, forecast accuracy — use specific pain points, not generic ones), then show our solution mapped to those pain points. Include: before/after comparison of their workflow with and without our tool, ROI calculator slide (time saved × team size), case study of a similar manufacturing client with specific results, implementation timeline, and pricing options slide. Tone: consultative, not salesy. Use ‘you’ more than ‘we.’ Every feature slide must answer the implicit question ‘so what?’ for their specific context.”

Why this works: It’s prospect-centric. It names the prospect’s situation, maps features to pain points, and demands the “so what” translation that separates consultative selling from feature-dumping.

Key prompt elements for sales decks:

  • Define the prospect persona and current situation
  • Mandate pain-point-to-solution mapping
  • Require evidence (case studies, ROI, testimonials)
  • Demand “you” language over “we” language

Scenario 4: Training and Onboarding

Training presentations need to teach, not just inform. Structure, repetition, and assessment matter.

Weak prompt: “Create a training presentation on cybersecurity best practices.”

Strong prompt: “You are an IT trainer delivering a 45-minute session to non-technical employees. Goal: after this session, employees can identify phishing emails, create strong passwords, and report security incidents correctly. Create a 15-slide deck. Structure: start with a real-world story of a breach caused by employee error (make it visceral, not abstract), then three learning modules (phishing, passwords, reporting) — each module follows the pattern: concept → real example → red flags → practice exercise. End with a 5-question knowledge check. Tone: approachable, not scary. Use analogies (password strength = bike lock quality). Avoid jargon. Every slide with more than 3 bullet points needs redesign — this is for learning, not reading.”

Why this works: It uses learning design principles (story hook, module structure, practice, assessment). It specifies the format constraint (max 3 bullets) that prevents the AI from producing reading-slides instead of teaching-slides.

Key prompt elements for training:

  • Define learning objectives as outcome statements (“can identify,” not “understands”)
  • Mandate examples and practice
  • Set bullet-point limits — training slides must be visual/interactive
  • Demand accessible language for the audience’s technical level

Scenario 5: The Conference Keynote

Keynotes are theater. They’re about vision, emotion, and memorable moments — not bullet points and dense data.

Weak prompt: “Create a keynote presentation about the future of AI.”

Strong prompt: “You are a CEO delivering a 20-minute conference keynote at a tech industry event. Audience: 2,000 professionals — curious but skeptical about AI hype. Goal: shift their perception from ‘AI is threatening’ to ‘AI is an opportunity if we shape it.’ Create a 15-slide deck. Structure: open with a counterintuitive hook (a surprising statistic or provocative question), three main sections (where AI is now, where it’s going, how to prepare), one ‘hero moment’ slide — a single bold statement or visual with no supporting text, close with a call to action that’s personal and specific. Maximum 30 words per slide. Heavy use of imagery — describe the desired image for each slide. Tone: visionary but grounded, personal (use ‘I believe’ not ‘it is believed’). No bullet points. No charts. This is a keynote, not a report.”

Why this works: It understands the keynote format — low text, high imagery, emotional arc, hero moment, call to action. It specifies the word limit and bans bullet points, which prevents the AI from defaulting to business-report mode.

Key prompt elements for keynotes:

  • Set strict word-per-slide limits (30 is generous for keynotes)
  • Demand visual descriptions — “describe the image” alongside the text
  • Ban standard presentation crutches (bullet points, charts, dense text)
  • Specify emotional arc, not just content outline

Scenario 6: The Internal Proposal

You need buy-in from your own organization. Different audience, different dynamics.

Weak prompt: “Create a presentation proposing a new project management tool.”

Strong prompt: “You are a department head proposing a tool change to your COO and finance director. They’re risk-averse and budget-conscious. Goal: get approval for a 3-month pilot. Create a 7-slide deck. Structure: current pain quantified (hours lost to tool friction × hourly cost), proposed solution with 3 options (not just one — shows you evaluated alternatives), recommendation with rationale, pilot scope and success metrics, cost breakdown including migration cost, and risk mitigation plan. Tone: pragmatic, data-driven, respectful of budget concerns. Acknowledge the downside of change upfront — don’t pretend there’s no cost. Include a ‘if this doesn’t work’ exit strategy.”

Why this works: It acknowledges internal organizational dynamics — the need to show alternatives, mitigate risk, respect budget, and provide exit strategies. It turns the proposal from “here’s what I want” to “here’s the thoughtful analysis.”

Key prompt elements for internal proposals:

  • Acknowledge the decision-makers’ concerns explicitly
  • Require alternatives comparison (option A, B, C with recommendation)
  • Include risk/exit planning — internal audiences want safety nets
  • Quantify everything — internal decisions run on numbers

The Universal Prompt Template

For any scenario, here’s a fill-in-the-blanks template:

You are a [ROLE] presenting to [AUDIENCE] in a [CONTEXT]. The audience cares about [WHAT THEY CARE ABOUT]. After this presentation, I want them to [DESIRED OUTCOME]. Create a [SLIDE COUNT]-slide [PRESENTATION TYPE]. Structure: [KEY SLIDES]. Tone: [TONE DESCRIPTORS]. Constraints: [NON-NEGOTIABLES]. Use [SPECIFIC ELEMENTS] — I’ll replace placeholders with real data.

The more specific you are about audience, outcome, structure, and constraints, the closer the AI gets to something actually usable. Vague in, vague out. Specific in, significantly better out.