FOR INVESTORS
Checking an AI product
We find out what stands behind the words “artificial intelligence” in a company’s pitch: its own technology, its own data or a layer on top of someone else’s model.
Discuss the task ↗Why this is a separate review
Connecting a ready-made model from OpenAI, Google or Anthropic to a product takes a few days, so almost every startup now calls itself an AI company. What matters to an investor is different: what stays with the company if the model provider raises prices, changes its terms or launches the same product itself.
What we check
- What the product does itself and what it takes from an external model
- Which data the company owns and whether it has the right to train models on it
- What it costs to serve one client and how that cost grows with sales
- Whether clients really pay for the AI features or the features exist only in the pitch deck
Available on its own or as part of a full company review. We give the timeline after the first conversation.
How the work goes
From a demo to a conclusion.
- 01
Demo
We look at the product working on real data, not on prepared examples.
- 02
Talking to developers
We work out how the system is built and what it depends on.
- 03
Numbers
We calculate model and infrastructure costs per client.
- 04
Conclusion
We write down what in the product is real, where the risks are and what to put in the deal.