Most failed projects were not built badly. They solved a problem few people cared enough about. The cheapest time to learn that is before you invest in a full product.
AI helps by making the test fast: you can sharpen the idea, draft the pitch and publish a test page in an afternoon, then spend your energy on talking to customers.
Validate before you build big: write down what must be true for the idea to work, put a simple page in front of the right people, and see whether they take a real action such as leaving a number or paying a deposit.
Write your assumptions down
Turn the idea into statements that can be wrong: who the customer is, what problem they have today, how they solve it now and why they would switch. Ask the AI to challenge each statement and point out what you are taking for granted.
- Who exactly is the customer?
- What do they do today instead?
- Why would they pay or switch?
Build a test page, not a product
A single page with a clear promise, a few benefits and one action is enough. The action could be a waiting list, a booking request or a pre-order message. The goal is to see behaviour, not to look polished.
Measure actions, not compliments
Friends saying it sounds great tells you little. Count the people who sign up, reply or agree to a paid pilot. Decide in advance what result would make you continue, change direction or stop, so you are not tempted to read weak results kindly.
- Messages or calls from people you do not know.
- Sign-ups from the specific group you target.
- Willingness to commit something, even a small deposit.
Use AI for research, but verify it
AI is good at brainstorming competitors' angles, customer questions and objections. It can also be wrong or outdated, so treat its market claims as leads to check with real people and real sources.
Example prompt to Uttra: "Create a one-page site for a home meal-prep service in Jeddah with a clear promise, three benefits and a form to join a waiting list. Add a question asking what meals they want most."