The pattern
- Wave 1 (2023): wrappers on GPT-3.5. Funded on vibes. Most are dead or pivoting.
- Wave 2 (2024): "AI-native" products with real retention data. Some are growing.
- Wave 3 (2025): vertical AI in regulated industries. Still early.
Why wrappers fail
The API is not a moat. If your differentiation is "we call OpenAI and add a nice UI," you lose when OpenAI ships that UI. Most 2023 AI startups were wrappers.
What survives
- Proprietary data (clinical records, legal documents, financial history)
- Distribution (existing enterprise contracts, channel partnerships)
- Workflow embedding (the tool sits in a process that is costly to change)
The takeaway
Survive long enough to accumulate one of those three. If you cannot explain which one you are building toward, your VC is asking the same question.