Using the AI

Writing effective prompts for better slides

5 min read

The quality of your prompt directly determines the quality of your slides. This article breaks down what makes a great prompt and shows you patterns you can reuse.

The four ingredients of a strong prompt

  1. Topic — What is this presentation about?
  2. Audience — Who will be in the room (or reading the deck)?
  3. Goal — What action or belief do you want the audience to leave with?
  4. Constraints — Slide count, tone, any must-include sections?

You don't always need all four, but the more you include, the less the AI has to guess.

Example: weak vs. strong prompt

Note

Weak: "Make a presentation about our company."

Tip

Strong: "Create a 12-slide company overview for new enterprise sales reps. Cover: who we are, our core product, the problem we solve, key differentiators, customer case studies (3), pricing tiers, and next steps. Tone: professional and confident. Avoid jargon."

Specifying tone

Tone words the AI understands well: formal, conversational, persuasive, inspirational, data-driven, educational, executive, technical, storytelling, minimalist. Combine them: "data-driven but accessible to a non-technical audience."

Asking for specific slide types

You can name slide types in your prompt and the AI will generate them: "Include a before/after comparison slide," "Add a timeline slide for our roadmap," "Finish with a Q&A / Thank You slide."

Iterating with the chat panel

After generation, use the chat panel to refine. Select a slide, then type a specific instruction. Examples:

  • "Make the bullet points shorter — max 8 words each."
  • "Replace the generic stats with placeholders I can fill in later."
  • "Rewrite this slide to focus on ROI, not features."
  • "Add a speaker note explaining the key takeaway for slide 4."

Using context files to improve accuracy

The single biggest improvement you can make to prompt quality is attaching a context file. Upload a product brief, annual report, or research document and the AI uses real data instead of hallucinating numbers or generic statements.