The most expensive mistake in startups isn’t a bad hire or a missed fundraise. It’s building something nobody wants.
I’ve done it. Most founders have. You spend 4 months building, launch to silence, and spend the next 2 months rationalising why the market isn’t ready yet.
The market was ready. Your idea just wasn’t right.
Here’s the validation framework I now run before writing a single line of code — and the AI prompts that make it faster.
The Validation Principle
You’re not trying to prove your idea is good. You’re trying to find out as fast as possible if it’s bad.
That’s the mindset shift. Most founders validate to confirm their bias. They talk to people who like them, ask leading questions, and interpret ambiguous answers as green lights.
Real validation is adversarial. You’re looking for reasons not to build this.
Week 1: Define and Research
Day 1-2: Get precise about the problem
Vague problems produce vague validation. “Hiring is hard” is not a problem you can build a business around. “Founders at 10-50 person B2B SaaS companies spend 8+ hours per week on hiring admin and lose candidates to faster-moving competitors” — that’s a problem.
Use this prompt to sharpen yours:
“I think [target customer] has a problem with [vague problem description]. Help me make this more specific. What is the exact situation they’re in when the problem occurs? What does it cost them in time, money, or opportunity? What do they currently do about it and why is that insufficient? Give me 3 increasingly specific versions of this problem statement.”
Day 3-5: Map the competitive landscape
“I’m building a solution for [specific problem] targeting [ICP]. Help me map the competitive landscape. Who are the direct competitors? Who are the indirect competitors (including ‘do nothing’ and ‘build it themselves’)? For each, what is their main weakness — and is that weakness something my target customer actually cares about? What would need to be true for someone to switch from their current solution to mine?”
Week 2: Talk to Real People
Day 8-10: Run discovery interviews
Aim for 10 conversations in 3 days. That sounds like a lot — it’s not if you use warm outreach and keep calls to 20 minutes.
Use this prompt to build your outreach message:
“Write a cold outreach message to a [ICP] asking for a 20-minute conversation about [problem area]. I am not selling anything — I am researching whether this problem is worth solving. Tone: direct and respectful of their time. Under 80 words. Include a specific reason why I’m reaching out to them in particular.”
Day 11-12: Synthesise what you heard
After your interviews, paste your notes into this prompt:
“Here are notes from [X] customer discovery interviews about [problem]: [paste notes]. Identify: (1) the 3 most common pain points mentioned unprompted, (2) the language people used to describe the problem — quote them directly, (3) what they’re currently doing about it and how satisfied they are, (4) any surprising insight I didn’t expect, (5) whether there’s a pattern suggesting willingness to pay. Be brutally honest about what the data does and doesn’t support.”
Day 13-14: The kill test
Before you decide to build, run this:
“Based on my validation research: [paste summary]. Make the strongest possible case for why I should NOT build this. What are the 3 biggest reasons this could fail? What would I need to believe to proceed — and how confident am I in each belief? What’s the one question I still can’t answer that matters most?”
If you can’t answer the kill test confidently, you need more validation — not a build sprint.
What Good Validation Looks Like
You’ve validated an idea when:
- At least 3 people described the problem in the same words without prompting
- At least 2 people asked when they could buy it
- At least 1 person offered to pay before it’s built
One person excited is noise. Three people excited is a signal. One person handing over money is proof.
Want the Full Product & Growth Prompt Library?
13 prompts covering validation, roadmap prioritisation, launch announcements, customer research, and more — free download.