Chapter 26 Key Takeaways: Presentations, Slides, and Visual Communication

Core Principles

Narrative before slides. The single most important discipline in AI-assisted presentation creation. If you ask AI to create slides before establishing the narrative, you'll get organized information that may or may not persuade anyone. Build SCQA first, then structure, then content.
AI accelerates creation; you determine purpose. AI can generate slide content, assertion titles, speaker notes, and visual recommendations rapidly. It cannot determine what your presentation is actually trying to accomplish or what this specific audience needs to hear. Those decisions are yours.
The audience's decision, not your analysis, determines the structure. Presentations organized around your analysis ("here's what we found") are almost always less effective than presentations organized around your audience's decisions ("here's what you need to decide"). Reframe from "what do I want to say?" to "what does this audience need to believe or do?"

Every compelling presentation has a Situation, Complication, Question, and Answer. The Complication is the most commonly underwritten element — the "but" or "however" that creates urgency. Without a well-defined Complication, presentations feel like status updates rather than calls to action.
The Answer should answer the Question. Sounds obvious, but many presentations have a Question the audience naturally asks ("what should we do about this?") and an Answer that doesn't directly respond to it. Test your SCQA by reading S → C → Q → A aloud. Does it flow logically?
Every piece of content should support the Answer. Slides that establish Situation, define Complication, or substantiate the Answer belong. Slides that are interesting but don't connect to the Answer are candidates for the appendix.

Assertion titles are the highest-leverage improvement you can make to most presentations. Converting "Market Analysis" to "The market is growing 40% annually, and we're taking share" makes the presentation argument visible at a glance.
Reading only the slide titles of a well-structured deck should convey the complete argument. If your title-only narrative doesn't hold together, the structure needs work.
The "one idea per slide" rule is an editing tool, not a design rule. When each slide must make exactly one claim, slides that aren't making a clear claim become visible — and can be fixed or cut.
Bullet points should be parallel and complete. Bullet fragments ("Market share," "Revenue," "Competition") communicate nothing. Complete assertions ("Market share is growing 8% annually") communicate a claim. AI defaults to fragments; prompt explicitly for complete thoughts.

Visual format should follow content type. Lists of parallel items: bullets. Data comparisons: charts. Processes and sequences: diagrams. Single powerful insights: large-type assertions. One powerful image: abstract concepts or emotional anchors. Default to bullets only when the content is genuinely a list.
Every chart needs a headline insight, not just labels. A chart with axes labeled and a legend is a data display. A chart with an assertion title that states the key insight is a persuasion tool. "Revenue growth accelerated in Q3" as a chart title is better than "Revenue by Quarter."
AI image prompts describe what to create or find; the actual image still requires your judgment. AI can generate precise briefs for images — what subject, composition, mood, and style would best serve the slide. Whether you then use AI image generation, stock photography, or custom design is a separate decision.
Real screenshots, real people, and real data make presentations feel authentic. AI-generated images work well for abstract concepts and visual metaphors. For product demonstrations, team slides, and customer evidence, authentic visual content is more credible.

Executive decks need to answer "so what does this mean for the business?" on every slide. Process detail, methodology, and comprehensive analysis belong in appendix or working documents. Executives need conclusions, not evidence trails.
The appendix is where you put content that serves "if they ask" rather than "what they need to know." A well-designed appendix allows you to go deep on any topic when questions arise, without cluttering the main narrative.
Adaptation for different audiences is primarily about emphasis and omission, not different facts. The same analysis, correctly emphasized, communicates differently to technical teams, executive sponsors, external customers, and investors. AI can restructure emphasis quickly when given specific audience context.

Q&A performance is the primary driver of post-presentation credibility assessments. Preparation for the questions you'll receive is as important as preparation of the slides. The most predictable questions are the ones you're most likely to be unprepared for because you've stopped noticing them.
Speaker notes should expand on the slide, not repeat it. Notes that say "As you can see on this slide..." are redundant. Notes that say "The reason this matters is..." are useful.
Generating likely questions before presentation is also a structural audit. If you can't answer a predictable question, your presentation may be missing evidence, an addressed objection, or a clearly stated conclusion.

The "AI-looking" deck problem is solved by specificity. Generic stock imagery, symmetric three-column layouts, and committee-style language are the signature of AI defaults. Replace them with specific examples from your organization, real data, actual screenshots, and language that reflects how you actually talk about your work.
The "comprehensiveness trap" requires explicit constraint. AI defaults to thorough coverage. Effective presentations require ruthless selection. Explicitly telling AI to "force you to choose" or to "assume the audience has 90 seconds per slide" produces better results than asking for comprehensive coverage and then editing.
Missing "so what" is a structuring failure, not a content failure. When a slide has information but no insight — data without conclusion, analysis without recommendation — the problem is structural. The fix is usually an assertion title or a "therefore" sentence, not more content.

Use AI tools for first drafts, not finished products. AI presentation tools produce useful starting points. The quality of the final presentation depends on how much human editing, context-addition, and judgment is applied after the AI draft.
The right tool depends on your organization's existing ecosystem. Microsoft 365 organizations should default to PowerPoint Copilot for the seamless document-to-presentation workflow. Google Workspace organizations should default to Gemini in Slides. Platform-agnostic teams have the most to gain from Gamma.app's content generation capabilities.
Structure first, then use AI tools. The worst workflow: open Gamma, type a topic, accept the structure. The best workflow: develop SCQA and narrative outline first, then use AI tools to generate content for an established structure.