Memorisely
AI Product Design

Start with a PRD, not a wireframe

Why designers should write the product requirements before drawing a single screen.

Zander Whitehurst

10 min read

For years, my first move on any new feature was to open Figma and start drawing boxes. It felt productive. Within an hour there was something to look at, something to share and something to argue about. The trouble was that most of the arguing was about the boxes, not the problem.

These days I start somewhere else. Before I draw a single screen, I write a short product requirements document. A PRD. It sounds like a job for a product manager, and traditionally it was. But with AI tools like Claude Code able to turn a clear written brief into a working prototype, the PRD has quietly become one of the most powerful design tools we have.

Here is why I think designers should write the PRD first, what goes into one, and how to go from a page of text to something you can click.

Why wireframes first falls short

Wireframes are a great way to explore layout. They are a poor way to explore thinking. When you start with screens, you make dozens of decisions about structure and flow before you have agreed what the thing is for. Those decisions harden quickly, and every review after that becomes a conversation about where a button sits.

A PRD forces the harder questions to the front. What problem are we solving? For whom? How will we know it worked? Once those answers are written down, the screens tend to design themselves, or at least they argue a lot less.

Starting with a PRD moves the first conversation from boxes to the problem.

Why this matters more with AI

When you build with AI, the PRD becomes the prompt. Claude Code, Cursor and similar tools are only as good as the context you give them. Hand them a vague one liner and you get a generic interface. Hand them a clear PRD and you get something that reflects your actual users, goals and constraints.

The PRD also becomes the shared source of truth. Designers, engineers, product managers and agents all read the same document. When someone asks why a feature works the way it does, the answer is written down. When the agent drifts, you point it back at the PRD rather than rewriting your instructions from memory.

To make this concrete, I will use a running example throughout: invoice reminders for small design studios. A feature that helps studio owners chase unpaid invoices without awkward emails.

1. Problem

What problem are we solving? Describe the pain in plain language, from the user's point of view. Avoid jumping to solutions here. If you cannot explain the problem in two or three sentences, you probably do not understand it well enough yet.

Small studio owners lose time and confidence chasing late payments, and often avoid sending reminders because it feels uncomfortable.

2. Users

Who are we solving it for? Be specific. Name the primary user and any secondary ones, and note what they care about. A useful test is whether the description would change your design decisions. If it would not, make it sharper.

Owners of studios with one to ten people who handle their own invoicing, usually between client work, often on a laptop late in the day.

3. Goals

What outcome are we trying to achieve? Goals describe the change you want to see, for the user and for the business. Keep them few and clear. Two or three is plenty.

Help studio owners get paid on time without having to write or send reminder emails manually.

4. Features

What should the product do? List the capabilities, not the screens. This is where it is tempting to add everything, so be disciplined. Separate what is essential for the first version from what can wait.

Automatically send a friendly reminder when an invoice is three days overdue, with a tone the owner can choose and edit.

One clear feature line in the PRD is enough for an agent to build a sensible first settings panel.

5. User flows

How should people interact with it? Describe the key journeys as short sequences of steps. Include the happy path, then the moments where things go sideways. This is the section that does the most work when you hand it to an AI tool, because it describes behaviour rather than appearance.

The owner opens an invoice, switches on reminders, picks a tone, previews the email and saves. When the client pays, reminders stop automatically.

6. Constraints

What technical, design and business requirements exist? Constraints are gifts. They narrow the options and stop both people and agents from inventing things you cannot build or support. Mention your design system, platforms, integrations and anything legal or commercial.

Must use our existing design system components, work with our current email provider and never send more than three reminders per invoice.

7. Success metrics

How will we know it works? Pick signals you can actually observe. They do not need to be sophisticated. They need to tell you whether the goal is being met, so you can decide what to do next.

Fewer invoices sitting overdue for more than two weeks, and owners keeping reminders switched on after their first month.

Keep it short

A PRD is not a novel. The best ones I have written fit on a single page. Each section gets a few sentences or a short list, and that is it. Long documents do not get read by people, and they confuse agents by burying the important parts.

A few habits help:

  • Write in plain language. If a new team member could not follow it, rewrite it.
  • Cut anything you are unsure about. Put open questions in a separate list at the bottom rather than mixing them into the requirements.
  • Prefer examples to adjectives. "A friendly reminder" is vague. A sample sentence of the reminder is not.
  • Revise as you learn. The PRD is a living document. Update it when the prototype teaches you something.
A sample sentence in the PRD says far more than the adjective friendly ever could.

I once wrote a four page PRD for a fairly simple feature and felt very pleased with myself. Nobody read past page one, including, if I am honest, me. The next version was a single page and the whole team could quote it back to me by the end of the week.

From PRD to prototype

Once the PRD is written, the fun part begins. Open Claude Code in your project, paste in the PRD and ask it to build a rough working version of the feature using your existing components. Ask it to start with the main user flow and to leave visual polish for later. We walk through this step by step in the AI Prototyping in Claude Code workshop.

The PRD becomes the prompt, so the first build reflects real users, goals and constraints.

Then click through what it builds. You will immediately spot gaps. Maybe the flow does not handle a client who pays partially. Maybe the tone picker needs a preview. Add those to the PRD, then ask the agent to update the prototype. The document and the prototype improve together.

When the core flow feels right, that is the moment to bring in more careful design thinking. Take the tricky parts into Figma if you need to explore visually, refine the details, and bring the decisions back. By now you are designing on top of something real, guided by a document everyone agrees on.

The first time I tried this with our team, we went from a blank page to a clickable prototype in an afternoon. More importantly, the conversation in review was about whether we had solved the problem, not whether the button should be blue.

Try it on your next feature

Before you open Figma for your next feature, open a blank document instead. Write the seven sections, keep it to a page, and hand it to Claude Code. See what comes back. I think you will find that the clearer your thinking is on paper, the better everything after it becomes. Writing PRDs like this is a core part of our AI Product Management Bootcamp.

Keep reading

View all
View all

Optimize your AI for better experiences, not AI slop.

View all

AI Prototyping Skills

Prototype with AI using clean UI patterns and guardrails so flows feel product-grade, not generic template slop.

$99

AI UI Motion Skills

Motion and micro-interaction patterns that make AI-built UI feel designed, polished, and trustworthy in real products.

$99

Fix AI Slop Skills

Spot AI slop early and fix type, spacing, and components so your product looks intentional, not like a template.

$99

AI Responsive Design Skills

Responsive layout patterns for rails, sidebars, and breakpoints so AI-generated UI adapts cleanly from narrow to wide.

$99

AI Dark Mode Skills

Dark theme tokens, contrast, and surfaces that keep AI-generated UI readable and product-grade, not flat gray on black.

$99

AI Color Token Skills

Neutral and primary ramps with clear steps so AI palettes stay systematic, not a random set of accent hex codes.

$99
View all