99% of Beginner's AI Presentations Suck (5 steps to fix it in ChatGPT + Gamma)

Key Concepts

• AI Presentation Limitations: AI struggles with tasks outside its "frontier" of knowledge and data.
• Input-Output Loop: Effective AI use requires a feedback loop with human processing and refinement.
• Hypothesis Building: Forming an educated guess based on context, goals, and existing insights.
• Deep Research: Utilizing AI tools for in-depth market analysis and data gathering.
• Targeted Dig: Focusing AI research on specific gaps and questions identified in initial findings.
• Framework-Driven Outline: Structuring presentations using established frameworks (e.g., SCQA, STR) for clarity and impact.
• AI Presentation Tools: Leveraging AI for quick drafts (ChatGPT Agent) or polished designs (Gamma).
• Centaur vs. Cyborg Approach: Adapting AI integration based on task division or seamless collaboration.

Why AI Presentations Suck: The Harvard Study

The video begins by addressing the common frustration that AI-generated presentations often fall short of expectations. It references a Harvard study that highlights a crucial distinction:

• Productivity Boost: Consultants using AI were significantly more productive, completing 12.2% more tasks on average and 25.1% faster, with 40% higher quality results.
• The "Frontier" Problem: However, for tasks outside AI's "frontier" (areas where AI lacks sufficient knowledge and data), consultants using AI were 19 percentage points less likely to produce correct solutions.

The "frontier" refers to the uneven capabilities of AI. AI excels at tasks with ample data (e.g., generating images of cats) but struggles with complex, nuanced topics like market entry strategies.

The Five Steps to Great AI Presentations

The core of the video outlines a five-step process to create actionable, insightful, and generative AI presentations:

1. Build Your Hypothesis

• Context: Understand your company's background, capabilities, and past experiences.
• Goal: Define the specific objective of the presentation (e.g., deciding on a market to enter).
• Insight: Identify any existing hunches or preliminary findings.

The human element is crucial here. AI lacks the contextual awareness and strategic understanding to formulate a meaningful hypothesis. The presenter suggests spending 20-30 minutes on initial research (e.g., Google searches, reviewing proprietary data) to form an educated guess.

Example: For a market entry project in East Asia, initial research might reveal that South Korea aligns well with a luxury-focused company due to higher per capita spending.

2. Deep Research with AI

• Tool Setup: In ChatGPT, enable "Deep Research" (over "Agent Mode" for its clarifying questions) and use a "reasoning model" (starting with "O") for analysis.
• Prompt Engineering: Provide clear context, task definition, and limitations in your prompt.

Example Prompt: "We are working on a market entry project in East Asia for online shopping, focusing on luxury. Our goal is to understand if East Asia is an opportunity and, if so, which country to prioritize. Do a SWOT analysis of the key countries."

The key benefit of "Deep Research" is its ability to ask clarifying questions, which helps refine the project scope and identify key considerations.

3. Targeted Dig

This step involves analyzing the AI's output from Step 2, "sense-checking" the results against your existing knowledge, and identifying gaps or surprising findings.

• Sense Checking: Does the AI's analysis align with your initial research and understanding?
• Gap Identification: Are there any areas where the AI is missing crucial information or nuance?
• Cross-Functional Talking Points: Identify topics that would benefit from discussion with colleagues from different departments.

Example: If the AI recommends China as the top priority but your company has existing offices in East Asia (excluding China), this discrepancy warrants further investigation.

4. Framework-Driven Outline

Instead of directly asking AI to create a presentation, first structure the content using established frameworks.

• SCQA (Situation, Complication, Question, Answer): A classic framework for problem-solving presentations.
• STR (Surprise, Complication, Resolution): A framework for engaging storytelling.

Example Prompt: "Create a 10-page presentation outline based on the data we have. The first page is an executive summary using SCQA. Then, lay out a SWOT analysis of each market. Conclude with slides on why Korea is the right answer and the associated risks. Include key data with sources."

5. AI Presentation

Two options are presented for creating the final presentation:

• Quick and Dirty (ChatGPT Agent): Suitable for internal discussions where aesthetics are less important. ChatGPT Agent can generate a presentation with cited sources, but the design is basic.
  • Agent Features:
    • Activity Tracking: Monitor the AI's research process and logic.
    • Browser Takeover: Manually provide credentials for accessing proprietary data.
    • App Integration: Connect to other software (e.g., Gmail, Canva) for richer data access.
    • Scheduled Tasks: Automate recurring research and reporting tasks.
• Polished (Gamma): Ideal for presentations requiring a professional look and feel. Gamma excels at design but relies on the user to provide well-structured content.

Workflow for Gamma: Paste the presentation outline from Step 4 into Gamma and choose a design theme. Gamma will automatically generate visually appealing slides.

Centaur vs. Cyborg: The Human-AI Partnership

The video concludes by emphasizing the importance of combining human expertise with AI capabilities. It references the Harvard study's distinction between "centaur" and "cyborg" approaches:

• Centaur: Dividing tasks based on strengths (e.g., human provides context, AI conducts research).
• Cyborg: Seamless collaboration and intertwined workflows (e.g., human tweaks and refines AI's output).

The most effective approach involves integrating both strategies, leveraging AI for speed and data analysis while relying on human judgment for strategic thinking, contextual understanding, and nuanced decision-making.

Synthesis/Conclusion

The key takeaway is that AI presentations are only as good as the process behind them. By following a structured approach that combines human expertise with AI tools, knowledge workers can overcome the limitations of AI and create presentations that are truly insightful, actionable, and impactful. The five-step process outlined in the video provides a practical framework for achieving this goal, emphasizing the importance of hypothesis building, targeted research, framework-driven outlining, and strategic use of AI presentation tools.