A slick demo tells you people might want the product. It tells you nothing about whether the company gets to keep selling it, or keep any money when it does. Here's what I'd want to know first. This is how I think, not investment advice.
What you'll have by the end: a checklist of questions and red flags you can run on any AI company before you get excited.
The questions I'd insist on
Whose model is it built on?
Many AI startups run on someone else's model (OpenAI, Anthropic, Google). That's fine, but it means their costs and even their product can change overnight. I want to know how exposed they are.
Does it make money per customer?
What does it cost to serve one user versus what they pay? If it loses money on each customer, growth makes the problem bigger, not smaller. (See the margins guide.)
Why would a customer stay?
What stops them leaving for the next tool, or for the model provider doing it directly? Data, workflow, brand, switching cost, something has to hold them.
Is the growth real or bought?
Subsidised credits and free trials can inflate usage. I want to see whether people keep paying once the discounts end.
Is there a legal question hanging over it?
Training data, copyright, regulation. A lawsuit can outweigh great product-market fit, as the Suno case shows.
The red flags
- Every metric is a "total users" number and none is about paying, retained customers.
- The pitch is all model quality and no answer on cost-to-serve.
- "We're the AI for X" with nothing that stops the model provider being the AI for X next month.
- Big valuation, vague revenue, and a lot of "reportedly".
Assume the product is good, people clearly like it. Spend your energy on the questions that decide whether it's a good business. That's where the risk actually lives.
Then run the full framework on a real company with the research a business guide.