How should founders pitch AI-native startups to investors?

Founders pitching AI-native startups today need to lead with defensibility, not capability. The era of impressing investors by demoing a GPT wrapper is over — sophisticated LPs and VCs now demand a clear answer to one question before anything else: why can't OpenAI, Google, or Anthropic ship this in six months? The strongest pitches in 2025 center on proprietary data moats, workflow lock-in, and compounding network effects that make the product structurally harder to displace over time.

The current funding climate rewards specificity. According to reporting from The Information this week, early-stage AI deals are increasingly bifurcating: infrastructure and foundation model plays are attracting massive rounds from a handful of mega-funds, while application-layer startups are being scrutinized far more rigorously on unit economics and retention. Investors want to see evidence of stickiness — DAU/MAU ratios, expansion revenue curves, and proof that users are integrating the product into mission-critical workflows rather than experimenting casually. Founders who walk in with cohort data showing 90-day retention above 40% are cutting through the noise.

a16z's recent guidance to portfolio founders, circulated through its Speedrun program, emphasizes framing AI-native products around "10x workflow transformation" rather than incremental automation. The pitch narrative that resonates is one where the founder can articulate a before/after for a specific professional persona — not "we use AI to summarize documents," but "a mid-market CFO closes her books three days faster and catches 30% more anomalies." Meanwhile, Y Combinator's latest batch demo day coverage from TechCrunch highlighted that the standout pitches shared a common structure: a crisp problem statement, a live product demonstration with real customer data, and a founder who could speak fluently to gross margin trajectory as AI inference costs decline.

The next signal to watch is how investors respond to "AI agent" companies as they move from demos to deployment. The critical question heading into Q3 2025 is whether agentic systems can demonstrate reliable, auditable outputs at enterprise scale — because that's the moment application-layer AI startups either graduate into durable businesses or get commoditized by the platforms they're built on. Founders who get ahead of that narrative now, with early enterprise case studies and clear reliability benchmarks, will be the ones closing rounds on their terms.

— Verge