Using AI to Create Presentations That Don't Suck

The Illusion of the One-Click Presentation

We've all been there. A deadline looms, and you have a dense report, scattered notes, or messy meeting transcripts. The promise of AI is irresistible: paste the text, type "make slides," and watch a professional-looking deck appear. The fonts are clean. The layouts are balanced. It looks, on the surface, like work is done. Yet, when you step back to review it, a hollow feeling creeps in. The slides list facts but lack a throughline. A minor detail gets undue spotlight. Your most powerful conclusion is tucked away on a cluttered slide near the end. The presentation has form, but it's missing its soul—a clear, compelling argument.

This happens because "create a presentation" is not one job. It's a cascade of interconnected decisions. The AI, given a single command, must make them all simultaneously and instantly. It must guess your core message, decide what the audience already knows, choose an opening hook, sequence the points logically, select supporting evidence, and determine what to leave out. When it guesses wrong on the foundational pieces—the 'why' and the 'story'—the entire structure suffers, no matter how pretty the slides are. You're left with a polished shell that fails to persuade.

The fix is not to abandon AI, but to change how you work with it. Stop treating it as a replacement for your strategic thinking. Start treating it as a powerful assistant that needs clear, phased instructions. The key is to separate the act of building the argument from the act of building the slides. This two-stage method ensures you own the strategy, while AI handles the heavy lifting of structure and execution.

A Better Way: The Two-Stage Workflow

Instead of asking for a finished product, break the process into two distinct phases with a clear handoff point. This method enforces a crucial discipline: you must approve the message before a single slide is designed.

Think of it like building a house. You wouldn't let a contractor start pouring concrete before you and an architect agreed on the blueprints. Stage one is the architectural plan—the argument, the flow, the supporting evidence. Stage two is the construction documents—the detailed slide-by-slide plan that brings the blueprint to life. Only after both are signed off do you worry about the paint colors and furniture, which in our analogy is the final visual design.

This approach flips the dynamic. You move from a passive recipient of AI's best guess to an active director guiding a process. The AI's role shifts from an oracle to a collaborator. You provide the source material and the strategic intent; it provides analysis, options, and structured drafts. The result is a presentation that is both strategically sound and efficiently produced.

Stage 1: Architect the Argument

This stage is purely strategic and happens entirely in a text document. Your goal is to produce a one-page blueprint that defines the presentation's intellectual core.

Start by giving your AI tool all the relevant source material. Then, give it a prompt focused on analysis, not creation. A good prompt instructs the AI to:

• Extract the Core Thesis: Based on this material, what is the single, most persuasive claim we can make?
• Identify Key Pillars: What three to four essential points must the audience believe to accept that thesis?
• Assign the Best Evidence: For each pillar, what specific data, quotes, or examples from the sources provide the strongest support?
• Propose a Narrative Arc: What is the most logical and engaging order to present these pillars? Should we start with a problem, reveal a surprise data point, or build a chronological case?

Here's the difference between a weak and strong Stage 1 prompt. Weak: "Summarize this market analysis into a presentation." Strong: "Analyze the attached market analysis report. Propose a single, actionable argument we should present to the sales team. Then, outline the three key reasons this argument is true, citing the most convincing statistics from pages 3, 7, and 12 for each reason. Finally, recommend a presentation sequence that will first capture attention and then build conviction."

The output is your blueprint: a proposed thesis statement, a list of points with attached evidence, and a suggested flow. Now, your critical work begins. Review this document ruthlessly. Is the argument bold and clear? Does a supporting point feel weak? Does the flow have a natural build? This is your chance to debate, edit, and refine. You can modify the text directly or provide feedback like, "The argument is too cautious. Make it more provocative based on the data on page 8," and have the AI revise it. Do not move forward until this one-page blueprint perfectly captures the story you want to tell.

Stage 2: Draft the Slide Plan

With the approved argument blueprint in hand, you now move to tactical execution. The goal here is to translate the strategy into a concrete presentation outline.

Provide the AI with two inputs: your original source material and the finalized, human-approved blueprint from Stage 1. Your new prompt is now explicitly instructional. It might say: "Using the approved argument document (attached) as the strict guide, create a detailed slide-by-slide plan for a 12-minute presentation. For each proposed slide, please specify: 1) A working title that states a claim or poses a key question, 2) Bullet points for core content or speaker notes that elaborate on the point, and 3) A specific instruction for evidence, such as 'Feature the bar chart comparing Q2 and Q4 results from the report.' Ensure the overall sequence follows the narrative arc we approved."

The AI will generate something like a script or storyboard:

1. Slide 1: Title: "Rethinking Our Client Onboarding: From Friction to Flow." Content: Opening hook about client retention challenges. Evidence: Reference anecdote from support logs.
2. Slide 2: Title: "Our Core Finding: Complexity is the Primary Drop-off Driver." Content: State the central thesis. Evidence: Include the data table showing a 70% correlation between step count and drop-off.
3. Slide 3: Title: "Pillar 1: The Five-Step Mandatory Setup." Content: Detail the first key problem. Evidence: Use the screenshot of the current setup form.

This slide plan is your second quality gate. You are checking that the elegant argument survived the translation into a practical speaking guide. Scrutinize each slide. Does its title actively advance the thesis? Is the assigned evidence the most powerful one for that point? Most importantly, hunt for "zombie slides"—slides that exist only because a piece of information was in the source material, not because it serves the core argument. A fascinating but tangential statistic must be cut. Every slide must earn its place by directly supporting the blueprint.

The Story Spine Test: Your Final Pre-Flight Check

Before you invest time in visual design, perform this simple, revealing test. Take the slide plan and read aloud only the proposed title of each slide, in sequence. Ignore all bullet points and evidence notes. Listen to the story these titles tell.

Does it sound like a logical, persuasive narrative? Or does it sound like a generic table of contents? Contrast two examples:

Generic List (Weak Argument): Introduction, Project Background, Methodology, Data Findings, Analysis, Conclusion, Next Steps.

Claim-Based Narrative (Strong Argument): Customer Feedback Signals a Hidden Problem, Our Data Reveals the Bottleneck is Shipping, Redesigning the Logistics Hub Cut Delivery Time by 30%, Here's How We Can Apply This Win Across the Network.

The first list informs. The second list persuades. If your slide titles don't form a compelling mini-story, the argument is still fuzzy. Go back and refine the Stage 1 blueprint or the Stage 2 translations. This test ensures the presentation has a strong backbone. When the spine is solid, adding visual muscle and detail becomes straightforward and effective.

Finding the Granularity Sweet Spot

This two-stage method is a practical application of a key principle in AI collaboration: task granularity. Granularity is the size of the work chunk you give to the AI. Get it wrong, and the process is frustrating. Get it right, and it feels like a superpower.

If the granularity is too fine—you prompt for each slide individually—you become a micromanager, exhausting yourself with constant direction. If it's too coarse—you ask for the entire deck in one go—you become a passive critic, stuck fixing buried bad decisions in a seemingly finished product.

The sweet spot is in the middle. You delegate the analytical labor of sifting through content and proposing a coherent structure (a medium-sized chunk). You retain the executive authority to define the core message and narrative flow (the highest-value human work). Then, you delegate the translation of that blueprint into a detailed plan (another medium-sized chunk), again with your approval as a checkpoint.

This balanced collaboration plays to the strengths of both parties. AI excels at processing large volumes of information, recognizing patterns, and generating structured drafts at speed. You excel at judgment, nuance, strategic intent, and understanding what will resonate with a specific human audience. By combining these strengths in a structured process, you don't just get slides an AI made. You get a persuasive tool you built, with intelligent assistance. That's how you create presentations with AI that are not just competent, but compelling. And that is the definition of a presentation that truly doesn't suck.