Skip to content
Lenny's 🏦 Vault

Finding Success in AI Products: Data Flywheels and Crafted Workflows

by Peter Dang on June 22, 2025

Peter Dang's approach to product leadership combines strategic systems thinking with a focus on team composition and human-centered design. His experience building products at Facebook, Instagram, Uber, and OpenAI reveals consistent patterns for success.

Building Products That Scale (1→100)

  • Plan your chess moves out in advance - think before you act

  • Build systems that let you go sustainably faster

    • Sometimes you have to go slow to go fast
    • Take time to build the right architecture that can scale
    • Example: At Facebook, they carefully designed the entire sharing loop for News Feed, which has remained largely unchanged for 12+ years
    • Example: At Uber, they created scalable abstractions for pickup/dropoff that could work across venues, airports, and countries with different conditions

  • Measure everything from the beginning

    • "You wouldn't fly a plane without instruments, so why would you run your product without understanding how it's doing?"
    • Build a growth team early - not just to drive growth but to force rigor
    • Growth teams ask the right questions that reveal what you're not measuring
    • This creates a virtuous cycle of data-driven decision making

Five Types of Product Managers

  • Consumer PM: Half designer, half product person

    • Obsessed with details and craft
    • "This is three pixels off and it's driving me nuts"
    • Focused on user experience and delight

  • Growth PM: Half data scientist, half product person

    • Numbers-first mentality
    • Naturally skeptical - "Show me the data"
    • Runs experiments to prove hypotheses

  • Business/GM PM: Half MBA, half product person

    • Starts with the business model
    • Thinks about margins, opportunities, value creation
    • Understands incentives and marketplace dynamics

  • Platform PM: Deeply wired to build tools for others

    • Builds systems that make everyone else go faster
    • Often overlooked but critical for scaling

  • Research/AI PM: Half researcher, half engineer, half product person

    • Understands the technology deeply
    • Can bridge between research and product
    • Crucial for AI products where model capabilities matter

  • Everyone has a primary and secondary type

  • Teams need a balance of these different archetypes

  • Different businesses need different types of PMs

Building AI Products That Win

  • Data flywheels are critical

    • Models get good at whatever data you show them
    • Proprietary data creates defensibility
    • Build systems that collect unique data through usage
    • Example: Companies like Warp have unique data on which code suggestions users accept/reject

  • Product craft still matters

    • Distribution advantages can be overcome with superior product experience
    • Example: Granola competing against Google Meet, Microsoft Teams, and Zoom
    • Small product details that delight users can create word-of-mouth

  • Understand vertical workflows deeply

    • Solve specific problems in specific contexts
    • Integration into existing workflows is key

  • AI is not a magic wand

    • It's malleable based on the data it's trained on
    • Product builders must channel AI capabilities into something humans want to use
    • The hustle and elbow grease of builders will still be required

Building High-Performance Teams

  • Hire for complementary strengths

    • Think of your team as a product
    • Build a "team of Avengers" with different superpowers
    • Create healthy tension between different perspectives (e.g., growth vs. craft)

  • The 6-month autonomy test

    • "In six months, if I'm telling you what to do, I've hired the wrong person"
    • Sets high expectations for both the manager and the hire
    • Creates a meta-goal of calibration rather than just hitting OKRs
    • Forces managers to be selective and support properly

  • Prioritize growth mindset

    • Look for people who can take feedback without defensiveness
    • Ask: "Tell me about your biggest mistake and how it changed how you work"
    • Create an environment where people can be vulnerable about failures
    • This unlocks continuous improvement

  • Let people lean into their strengths

    • Help people create roles that match their unique abilities
    • Example: Creating the "Model Designer" role at OpenAI based on a team member's unique combination of technical depth and product taste
    • Have people write down what they're excited about to codify new roles

Effective Communication

  • The power of language shapes thought

    • Be intentional with the words you choose
    • Craft vision statements and goals with precision
    • The right language can change how people think about problems

  • The three-part communication loop

    • Say you're going to do the thing
    • Say that you're doing the thing
    • Say that you did the thing
    • This creates alignment, invites feedback, and ensures visibility

Career Development

  • Optimize for learning over prestige or compensation
  • Look for companies with unique insights about human behavior
  • Seek roles where you can apply your natural strengths
  • "If you move a tree, it dies. If you move a person, they thrive."
View full episode