LinkedInNoteEngineering Leadership

How should we build AI products?

Don't ask users what they want. Build a product with a point of view, create something that feels magical, and focus on real adoption instead of hype.

LinkedIn
July 20, 2025
Read time
2 min
Language
English
Engineering LeadershipJul 20, 2025English
  • "Don't ask users what they want. Build a product that sets direction."
  • "The best AI products don't feel mechanical. They feel magical."

Step-by-step strategy

  1. Build opinionated products: don't ask users what they want, set a direction.
  2. Deliver magical experiences: not mechanical, but surprising.
  3. Tool -> copilot -> autopilot: raise the level of automation gradually.
  4. 95% is company-building: focus on team, execution, and product.
  5. The AI-specific 5% matters at scale: data flywheels, trust, and UX.
  6. Build a data flywheel: user behavior makes the model smarter.
  7. Focus on real revenue: not vanity, but actual adoption, retention, and behavior change.
  8. Don't fear low early margins: token costs keep falling, scale is the answer.

https://lnkd.in/gYahhDZv

LinkedIn attachment 1