Most business ideas die from the same thing: building first and asking later. AI won’t guarantee success, but it lets you do in one afternoon the research that used to take weeks — and find the holes in your idea before spending a euro. This is the process we’d follow, step by step and with feet on the ground.
Step 1: stress-test the idea (devil’s advocate)
- The key prompt: “Act as a skeptical investor. My idea is [X]. Give me the 10 reasons it would fail and what would have to be true for it to work”. It hurts, but it saves months.
- Chase the problem, not the solution: ask the AI to describe the person suffering the problem you solve: who they are, how they solve it today, what it costs them. If that portrait isn’t convincing, bad sign.
- Beware the cheerleading: chatbots tend to encourage you. Explicitly ask for the opposite (“be brutally honest”) or alternate between two models to contrast.
Step 2: research the market and competition
- Competitive map in minutes: with AI search (Perplexity, ChatGPT with browsing) ask for “companies solving [problem] in my market, with their pricing model”. Verify the links: it’s research, not revealed truth.
- Listen to real pain: ask it to summarize common complaints in reviews and forums about existing solutions. Your differentiation lives there.
- The number that matters: are people already paying to solve this? A market with charging competitors is a better sign than a desert.
Step 3: validate with something real (not opinions)
- Build a smoke test in a day: a landing page explaining the offer (AI writes the copy) with an “I’m interested” button. Measured interest beats a hundred friendly opinions.
- Prepare real interviews: ask the AI for a discovery interview script (“Mom Test” style: questions about their life, not your idea) and practice with it playing a difficult customer.
- Simulate the business: “with these price and cost assumptions, how many customers do I need to break even?” — the AI builds the basic math and shows which variable kills the model.
Our take: what AI cannot validate
- AI processes the known; your customer is the unknown. No chatbot replaces ten real conversations with people who suffer the problem.
- Beware armchair validation: if all your “validation” happened inside a chat, you haven’t validated anything — you’ve planned. The real signal is someone willing to pay or give you their email.
- Where it truly shines: compressing weeks of desk work into hours, and forcing you to think rigorously before falling in love with your idea.
Frequently asked questions
Which AI should I use for this?
Any of the big ones works; ideally combine one with web search (for market and competition) and one strong at long reasoning (to stress the idea and build scenarios).
How long does validating like this take?
The desk research, one afternoon. Real validation (landing + interviews), a couple of weeks. Still infinitely cheaper than building blind.
Conclusion
Validate cheap, fail fast and only build what survives scrutiny. If the idea passes the filter, here’s how to make money with AI and how to sell AI agents to businesses to execute it.
