AI-focused investors have seen more pitches in the last two years than most VCs see in a decade, which means your "we're using AI to do X" framing won't land—they've heard it a thousand times. What these investors actually want to understand is your defensibility story: where your moat comes from when the underlying models are commoditizing fast. If you're building at the application layer, come prepared to talk about proprietary data flywheels, workflow lock-in, or vertical expertise that can't be replicated by a foundation model update. If you're building infrastructure, expect deep technical diligence on why your approach won't get absorbed into the major cloud providers' offerings within 18 months. The bar for technical credibility is high here—many of these investors have seen enough AI companies to spot hand-wavy architecture slides immediately.

​

One dynamic worth understanding: AI investors are increasingly bifurcated between those chasing the next foundation model play (massive rounds, long timelines, compute-intensive) and those focused on applied AI with faster paths to revenue. Know which type you're talking to before the meeting, because their return expectations and patience levels are completely different. The foundation model hunters want to see world-class research teams and are comfortable with years of negative unit economics; the applied AI folks want to see paying customers and gross margins that don't collapse when you scale. Also worth noting: many generalist funds have added "AI" to their thesis but lack the technical depth to evaluate infrastructure plays—these can be good partners for go-to-market-heavy application companies but may struggle to help you navigate model selection, build-vs-buy decisions, or hiring ML talent.