Facebook (SV-Asia Venture Hub 矽谷-亞洲新創投資討論社團)
businessai-engineering
Original source

AI 讓創業最難的事情變了 — NTU Demo Day in Silicon Valley

Summary

Observations from NTU students’ Demo Day in Silicon Valley after a 7-week program. Core thesis: AI has made building products easy, but the hard problems have shifted downstream.

Key Insights for AI Product Design

  1. “Who pays?” is the first question, not “how good is your AI?” — Judges barely asked about AI implementation; they focused on business viability
  2. User ≠ Buyer — Silver Memory (elderly AI companion): user is grandparent, decision-maker is child, payer might be healthcare/insurance institution
  3. Pain ≠ Business — A very painful problem doesn’t automatically equal a good business. Must find the economic buyer.
  4. Frequency matters — UniCUDA (university application platform): intense need but only once per year, $29/season. One-time use vs. daily recurring use have completely different business value.
  5. Better tech ≠ deeper moat — AI LinkedIn challenger asked “how do you beat LinkedIn?” Real moats are network effects, trust, distribution, proprietary data, and 20 years of user habits.
  6. Demand-pull > Tech-push — The strongest signal for early startups: someone with the problem came to YOU (not you building cool tech then looking for buyers). The EU Digital Product Passport compliance startup had a textile company approach them first.
  7. 0→1 is now cheap, 1→paying customer is not — AI coding/design/agents make MVPs trivial. But finding customers, building trust, distribution, and competing remain just as hard.
  8. The hardest question shifted — From “Can I build it?” to “Is what I built worth existing?”

The 6 Eternal Questions (regardless of AI)

  • Who pays?
  • How often do they use it?
  • Why you?
  • Why can’t others do it?
  • How do you find first customers?
  • Where is your moat?