Multi-Agentic Flow for Researching Winning Playbooks from $100m Entrepreneurs

multi-agentic flow for researching winning playbooks from $100m entrepreneurs

I'm scaling my agency to $100k/mo and want to find entrepreneurs who've done that before. I typically research them by watching videos and taking notes—which is valuable—but I wanted to see how far I could push AI using a few different tools. The output I want to learn is pretty much the same for any entrepreneur.

What I want to know:

• What niche did they start in
• What niche are they in now
• How long did it take them to get to where they are now
• What tactics did they use each year, and what amount of $$$ did they generate during that period
• What are the SPECIFIC daily actions they took (for example, rather than generic SEO & content planning, specific actions like 5× 1,000-1,500 word keyword-researched listicle-style blog posts plus outreach to high domain authority sites)
• If you can't find the specifics of the tactic in the YouTube video, go research using the web
• Timeline of their growth over time

By the way, in the AI/tech world this is known as structured JSON output—giving the AI a set of instructions to complete its job. (If you're not technical, feel free to gloss over this part.)

interface CreatorGrowthAnalysis {
  initialNiche: string; // The niche they started in
  currentNiche: string; // The niche they are in now
  timeToSuccess: string; // How long it took them to get to their current level
  yearlyTactics: YearlyTactics[]; // A breakdown of tactics and revenue by year
  specificDailyActions: string[]; // Granular daily actions they took
  researchRequired: boolean; // Indicates if additional research beyond the video was needed
  growthTimeline: GrowthEvent[]; // A timeline of their growth over time
}
interface YearlyTactics {
  year: number;
  tactics: string[]; // Specific strategies used during this period
  revenueGenerated: number; // Amount of $$$ generated that year
}
interface GrowthEvent {
  year: number;
  milestone: string; // Key events in their growth journey
}

Nerd stuff aside, I tried two different methods for this.

Method 1

I wanted to try out Google's new release from yesterday that lets you give AI Studio the ability to accept YouTube videos. However, I noticed it wouldn’t give me cited links. It was easier to get started because I just pasted the video link—but the output just wasn’t it.

Method 2 (the winner)

I used my AI-assisted terminal (warp.dev) and asked it to download the YouTube video (link of video) and turn it into a text transcript in the current folder. It did that in a few minutes, then I went into ChatGPT Playground and crafted a prompt to take in the transcript and return all the information I wanted.

I used o3-mini-high reasoning and wrote a function call for it to use the internet (though it may have just used its internal web search tool). The output was amazing—so good, in fact, that I think this could be a standalone product. It wouldn’t be a SaaS, but rather something people could visit to learn and grow.

Anyways, the prompt is here: https://platform.openai.com/playground/p/tgwCqUuzDs8a6ggqF8T83jzB?mode=chat I cover this live on my YouTube channel where you can follow along live... it'll go out tomorrow, Fri March 14, 2025. here