AI 스타트업 IR, 투심위가 「그래서 뭐가 남느냐」를 묻는 이유

Half of an AI team's deck is technical explanation. Yet the first question at the investment committee is usually this: “When the models get better, what is left of this company?”

Model performance is not a moat

Even six months ahead today, the gap disappears with the next release. Investors have been through this enough times that benchmark slides score worse than founders expect. A benchmark proves the technology works, not that the business works.

Models can be rented. What cannot be rented is the actual subject of the pitch.

What is left is usually one of three things

• Data, 서비스를 쓸수록 쌓이고, 남이 사올 수 없는 형태인가
• Workflow embedding, 고객사의 업무 흐름 안에 들어가 있어서 빼면 일이 멈추는가
• Switching cost, 옮기려면 무엇을 다시 세팅해야 하는가

With none of the three you have a good feature; with one of them clearly true you have a company. Put that sentence at the front of the deck and move the technology behind it, and the Q&A changes direction.

Do not hide the cost structure

Whether inference cost scales with revenue, or unit cost falls as customers grow, always comes out in due diligence. A team that raises it first reads very differently from one that starts calculating when asked.

In short

• Slide 1 — how people handle this today and what it costs
• Slide 2 — what shrinks and by how much, in numbers
• 3장 : 모델이 바뀌어도 남는 것, 데이터 · 접점 · 전환 비용 중 무엇인가
• Slide 4 — if revenue grows tenfold, what happens to cost

The technical explanation can wait until after those four slides.