Define What You're Building: Create PRDs with AI as Your Thinking Partner

In 1975, Frederick Brooks published The Mythical Man-Month and delivered a painful truth: adding more people to a late software project makes it later. His solution? Spend more time defining what you're building before you build it. "The hardest single part of building a software system," he wrote, "is deciding precisely what to build." Fifty years later, that insight hasn't aged a day.

Every product manager knows why. You open a fresh document, cursor blinking, ready to capture that brilliant idea. An hour later, you've written three sentences and deleted two of them. The problem isn't that you don't know your idea - it's that you know it too well to explain it simply. Your brain holds a rich, interconnected web of context, but the document demands linear clarity.

So you skip the document. You explain your idea in a meeting instead. The engineer explains it to another engineer. By the time code gets written, your vision has mutated into something you barely recognize. This isn't anyone's fault - it's what happens when ideas live only in conversations. The Product Requirements Document exists to break this chain, giving everyone a single source of truth.

But PRDs are snapshots, not monuments. They capture your current understanding of a problem and its solution. Once you start building, that understanding evolves - which means your PRD needs to evolve too. The document becomes a living record of your thinking, keeping everyone aligned as clarity emerges.

The challenge is getting that initial version right. You need enough structure to start building, but the process of creating it feels like pulling teeth. That's where AI changes the game - not by writing for you, but by workshopping with you until your thinking crystallizes into clear requirements.

Solution: Let AI workshop your requirements with you

Here's a quick walkthrough of how it works:

  1. Start with what you have. Paste your existing one-pager (from my previous skill), raw notes, or even nothing at all. The AI adapts to whatever input you provide.
  2. AI conducts the workshop. Instead of you prompting the AI, the AI prompts you. It asks focused questions to extract your understanding of the problem space, target users, and proposed solution.
  3. Define user stories together. The skill guides you through creating user stories with acceptance criteria. It suggests options, but you can always provide your own context if none of the suggestions fit.
  4. Prioritize for your first version. Not everything belongs in v1. The AI helps you identify what's essential to start building versus what can wait for later iterations.
  5. Get a structured PRD. The output is a well-formatted requirements document with user stories grouped by priority, business rationale, and UX requirements - ready for feedback or to feed into your vibe coding project.

For setup instructions and the full skill file, check out my GitHub repository.

Why Claude works best for this

The skill works with ChatGPT, Gemini, and Claude. I demonstrate it in Gemini in the video above, but I strongly recommend Claude for this particular workflow. Claude feels like a human thinking partner. When I thought I had full clarity on an idea, it exposed gaps I hadn't considered. It asks follow-up questions that actually matter and doesn't make up requirements I didn't ask for. For the first time, I get that slightly uncanny feeling of workshopping with a real colleague.

Resources