AI-powered discovery
Discovery is where most product ideas should die. If you can't answer basic questions clearly, the product isn't ready to build.
AI tools make discovery faster, not easier. You still need to think. But you can validate assumptions in hours instead of weeks.
The Goal
End discovery with a 1-2 page brief that answers:
- Who is this for? Specific persona, not "everyone"
- What problem does it solve? The job to be done
- Why will people pay? Value proposition in dollars
- What's the MVP? Minimum feature set that solves the problem
If you can't answer these clearly, kill the idea or keep iterating.
The Process
1. Problem Interview with AI
Use Claude or ChatGPT to simulate customer interviews. This sounds weird but works.
Prompt template:
Act as a [persona - e.g., "small business owner who manages
inventory manually"]. I'm going to ask you about your current
workflow and pain points. Respond based on realistic challenges
this persona faces.
Ask:
- What's your current workflow?
- What breaks first when volume increases?
- What workarounds have you tried?
- How much would you pay to solve this?
The AI won't give you perfect customer insight. But it will surface assumptions you haven't questioned.
2. Competitive Landscape Analysis
AI can scan and summarize competitors in minutes.
Prompt template:
Find 5-10 existing solutions for [problem space]. For each:
- Core features
- Pricing model
- Target customer
- Main differentiator
Look for gaps. Where do existing solutions fail? What use cases do they ignore?
More important: If there are no competitors, that's a red flag. Either the market is too small or the problem isn't real.
3. Feature Prioritization
List every possible feature. Then force-rank them.
AI helps by asking hard questions:
- Can this work without feature X?
- What's the minimum to demonstrate value?
- What can we ship in phase 2?
4. Technical Feasibility Check
Before you commit to building, validate the technical approach.
Ask AI:
- What's the standard stack for this type of product?
- What are the hard technical problems?
- What third-party services exist for X?
- What's the infrastructure cost at 100 users? 1000?
You're not designing the architecture. You're checking if there are landmines.
5. Write the Brief
The brief is your forcing function. If you can't write it clearly, you don't understand the product.
Tools
- Claude or ChatGPT - Simulated interviews, competitive analysis, feature prioritization
- Perplexity - Research on market size, competitors, technical approaches
- Notion or Google Docs - Brief writing and iteration
Red Flags
Kill the idea if:
- You can't define the target customer specifically
- The value proposition requires explanation beyond one sentence
- Your MVP has more than 10 features
- No existing solutions exist (usually means no market)
- You're solving a problem you've never experienced
Timeline
Discovery should take 1-3 days. If it's taking longer, you're overthinking it.
The brief doesn't need to be perfect. It needs to be clear enough to prototype against.
Most ideas die here. That's good. Better to kill bad ideas in discovery than in production.