Two years ago, "use AI to build your product" mostly meant autocomplete for code. Today, an AI model can help you interview customers, write a spec, generate a clickable prototype, and ship working software. The newest models from Anthropic, Google, and OpenAI are explicitly built for long, multi-step coding and knowledge work.
That's a real opportunity, especially for small teams and solo founders who have the expertise but not a ten-person engineering team. It also creates a new failure mode: building the wrong thing, very quickly.
The founders I see getting the most from AI treat it as a partner with enormous capacity and no judgment of its own. It will happily build whatever you describe. Your job is to make sure what you describe is worth building. Here's how I work with it at each stage.
Stage 1: Frame the problem before you touch a tool
Write one paragraph: who has the problem, what they do about it today, and what it costs them. Then ask an AI model to argue against you. "Here's the problem I think exists. What are the strongest reasons it might not be a real or urgent problem? Who else is already solving it?"
You're not asking for permission. You're stress-testing the idea before you invest in it. If the counterarguments are strong, that's information you got for free.
Stage 2: Let AI do the heavy lifting on research synthesis
Talk to real people — five to ten conversations is plenty to start. Record and transcribe them (with permission). Then use AI to pull out patterns: the words people use for the problem, what they've tried, what they'd pay to make go away.
The partnership works because each side does what it's good at. You have the conversations, notice the hesitation in someone's voice, and ask the follow-up question. AI reads every transcript with equal attention and doesn't forget the third interview by the time you reach the eighth.
Stage 3: Write the smallest useful spec
Ask AI to help you write a one-page spec for the first useful result — the smallest thing that solves the core problem for one type of customer. Have it list what's in scope, what's explicitly out, and how you'll know it worked.
Then cut it in half. AI tends to be generous with features. A good partner pushes back on that, and on this one point, the pushback has to come from you.
Stage 4: Prototype to learn, not to launch
This is where AI changes the economics most. You can go from spec to clickable prototype in an afternoon. Use that speed to test with real people before you build the real thing: show the prototype, ask them to complete the core task, and watch where they hesitate.
The rule I give founders: throw away your first prototype on purpose. Its job is to teach you what the second one should be.
Stage 5: Build with guardrails
When you do build, AI coding tools are now genuinely capable — but they work best inside a structure you set:
- Small, reviewable steps. Ask for one feature at a time and read what changed.
- Tests that describe the behavior you want. They keep the AI honest when it says something works.
- Clear permission boundaries. Decide what an AI agent may do on its own (edit code in a branch) and what needs your sign-off (anything touching production, payments, or customer data). The industry's recent focus on safety — including OpenAI shelving a model in September after tests showed it acting without authorization — is a good reminder to set those lines deliberately.
Stage 6: Measure the thing that matters
Once real people use it, the question shifts from "does it work?" to "does it help?" Ask AI to help you define one number tied to your first useful result and set up the simplest way to track it. Then review it weekly, together: you bring the context, it brings the analysis.
What the partnership looks like when it's working
You'll know you've got the balance right when AI is doing most of the typing and you're doing most of the deciding. If it feels like the reverse — you're typing prompts all day and the AI is choosing your roadmap — step back to Stage 1.
This is the same way we built Waymaker, and it's how we build products with clients. If you have an idea, or a product that isn't getting traction yet, we can help you scope the first useful version and get it moving.
Want this applied to your business? The Growth Fix is one fixed-price session to find where your customers drop off and fix the part that matters most.
