AI shifts moat to sales

Question / Claim

AI has made building products so easy that the main work for creating a million-dollar company is customer conversations, selling, and finding product–market fit.

Key Assumptions

• AI tools significantly reduce time/cost to build MVP-level software for many products.(high confidence)
• Distribution, sales, and customer insight are now the limiting factors more often than engineering execution.(high confidence)
• Most million-dollar businesses can be achieved with small teams if customer acquisition and retention are solved.(medium confidence)
• Competitive advantage is shifting away from proprietary code toward go-to-market and positioning.(medium confidence)

Evidence & Observations

• Observation: with AI assistance, small teams can prototype and ship faster than before, making it easier to reach a usable product quickly.(personal)
• Controlled experiment reported by GitHub: developers using GitHub Copilot completed a coding task ~55% faster than those without it (evidence that AI compresses MVP build time).(citation)
• Field experiment in a large call center found generative AI assistance increased worker productivity (notably improving outcomes for less-experienced workers), suggesting AI can shift bottlenecks from execution to workflow/design and customer-facing work.(citation)
• Study of knowledge-work tasks shows LLM tools can meaningfully change speed/quality—but benefits are uneven across tasks (“jagged frontier”), implying human judgment, positioning, and customer understanding remain critical even as building accelerates.(citation)
• Classic startup guidance emphasizes that early traction comes from direct user conversations and manual sales/onboarding (“do things that don’t scale”), aligning with the idea that go-to-market dominates once prototyping is cheap.(citation)

Open Uncertainties

• How much does this vary by industry (regulated sectors, hardware, deep tech) where building is still hard?
• Does AI also reduce sales/distribution costs enough to change the bottleneck again?
• What specific go-to-market channels are most leverageable for AI-native products right now?
• At what point does product quality/engineering become the bottleneck again (scale, reliability, security)?

Current Position