1. Intro & Introduction of Pat Grady 🎤

• Host Manny introduces Pat Grady, emphasizing that he is a renowned venture capitalist who has invested in major companies like Zoom, Okta, Notion, and HubSpot.
• "Pat, I'm so glad to have you here. He is one of the best investors and, rare as it is, a genuinely great human being."

2. AI Startups Are, at the End of the Day, About Building a Company 🏗️

• Pat references a meme about AI startups, saying "Building an AI company is ultimately the same as building any company."
• "Like the Scooby-Doo meme — pull off the mask of an AI company and you still find just a 'company' underneath."
• The core 95% is identical to any other company: hiring the right people, articulating a clear mission, rallying the team — basic management fundamentals matter most.
• The differentiating 5%: AI-specific factors exist, but most problems still come down to people problems.

3. Two Keys to AI Startup Success: Trust and Persistence 🔑

3-1. Trust

• "People don't trust AI yet. They don't trust your agent. You have to earn that trust."
• Ways to build trust:
  • Show the work: Make it transparent what the agent is doing and how it arrived at an answer.
  • Cite sources: Clearly provide the reasoning and sources behind results so users can verify for themselves.
  • "Like elementary school — it's not enough to get the right answer; you have to show your work."
• Top-down distribution model: As in enterprise sales, human-to-human trust relationships matter.

3-2. Persistence

• "People who are negative about AI are people who haven't tried hard enough. Getting to 80% is easy — the real magic is in closing that last 20%."
• "If you stop at the 80/20 experience, you fail. You have to push all the way through and build something worthy of trust before you can succeed."

4. Portfolio Examples That Build Trust Well 🏆

• Day AI (AI-powered CRM): A company founded by Christopher O'Donnell that "ignores vanity metrics like adoption, engagement, and revenue — and focuses solely on product trustworthiness."
• "When a pipeline is auto-generated, you have to give enough context explaining why — otherwise users get anxious."
• Harvey (legal AI): Built trust with the most demanding law firms, operating on the strategy that "if you earn trust at the top, everything below naturally follows."
• Open Evidence (medical AI): Started by "saying 'I don't know' when it doesn't know" and gradually raised accuracy to build trust.

5. AI Startup Characteristics: Small Teams, High Productivity, Fast Growth 🚀

• "AI companies achieve 10–100x productivity gains even with a small headcount, by putting AI to work aggressively."
• "Past tech waves were slow on distribution, but the internet is already laid — you can scale fast."
• "Before, you had to build a product and grow a sales force to grow. Now a small team can scale quickly."

6. 'Vibes Revenue' vs. Real Revenue ⚡

• "Since ChatGPT, everyone is trying AI — but the real problem is 'vibes revenue.'"
• "If you only deliver the magic moment (the 80/20), users leave quickly. You have to solve the problem all the way through to generate real revenue."
• "Metrics like one-month retention get a lot of attention, but what truly matters is sustained engagement and return visits."

7. New Core Metrics AI Startups Should Track 📊

1. Consumer internet metrics: DAU/MAU, daily/weekly retention — focus on user engagement and return visits.
   • "Even with just five days of data, you can see how many people who came on day one return on day two, and how many stay through day three."
2. Data Flywheel: "More usage → more data → better model → more usage — a virtuous cycle you actually have to prove."
   • "Most founders talk about data flywheels, but only 1% actually execute on them."
   • "You have to show logic and evidence. Handwaving doesn't cut it."
   • Example: Noom (a nutrition app) went from one coach managing 20 clients to managing 20,000 after integrating AI.
3. Margin (profitability) is not urgent right now: "Low margins today are fine if they can improve over the long run."
   • "Token prices have dropped 99%, and data-privacy demands have eased, so cost structures are improving fast."

8. The Real Moat of an AI Startup Is the Founder 🧑‍💼

• "The biggest moat is the founder themselves. You have to be able to say, 'I am the moat. I won't stop until I've succeeded.'"
• Companies like Amazon and DoorDash wouldn't exist without the founder's relentless persistence, he notes.
• "As an investor, ultimately betting on the founder is the single most important investment thesis."

9. AI Service Pricing: Input vs. Output and New Units of Value 💸

• "Outcome-based pricing is likely to become the standard for AI services, though it may vary by market."
• "Expo (fintech AI) invented a new unit called 'hacker hours,' expanding existing labor-based service markets."
• "Just as AWS created the server unit and Snowflake created the capacity unit, AI can price by units of human work."
• "You have to make both input (what the agent does) and output (business value) transparent, and explain it in units the customer understands."

10. Areas They Don't Invest In & the Road Ahead 🔮

• "Foundation models are closer to MongoDB (a database) than to AWS. They can be tens-of-billions businesses, but not hundreds-of-billions or trillion-dollar businesses."
• "The real value ultimately lives at the application layer. Domain-specific AI (legal, medical, CRM, etc.) will remain as independent companies."
• "The key question going forward is where the foundation model's domain ends and where the ecosystem above it begins."

11. Closing & Final Thoughts 🌟

• "Thank you so much for today. I'm grateful for your support and friendship."
• Pat: "Congrats on the launch, and thanks for having me."

Key Takeaways

• AI startup = still just a company
• Trust and Persistence
• Transparency in the work process
• Actually executing on the data flywheel
• The founder (the person) is the biggest moat
• Input/output-based pricing and new units of value
• Fast growth, high productivity, small teams
• Vibes revenue vs. real revenue
• Consumer internet metrics, retention, engagement
• The limits of foundation models and the value of the application layer

"Building an AI startup is ultimately the same as building any company. Earn trust, push through relentlessly — and remember, the real moat is you." 🚀