From PRD to Prototype
Test ideas, not perfection. Four steps to turn your PRD into a clickable prototype with AI.
Agenda
Segway: is this an MVP? A prototype? First-gen?
Have a think before I tell you the story. What would you call it?
[DAVE: your live framing / audience poll for this question]
Segway: solution first, product later
- Raised more than $100M in 2003
- 7 years and $100M later
- Sold to an investment company for $9M in 2013
- Segway-Ninebot: kick scooters
- Became profitable in 2015
- Sold to Ninebot (Shenzhen) for $75M+
Break up the validation steps
The traditional way
Cost: time + $$$ + opportunity cost, all spent before you learn anything
Faster and cheaper
Validate the problem first, then validate the solution. Two smaller bets instead of one big one.
There are multiple ways to solve a problem
And then: mobile app? Tablet app? Chat bot? Website? All of the above?
Each one is a guess until a customer tells you otherwise. A prototype is the cheapest way to ask.
The goal isn't to make a wireframe, prototype, or MVP. Those are just steps.
The real goal is to find paying customers as fast as possible.
Wireframe, prototype, PoC, MVP: what's the difference?
| Step | What it is | What it tests |
|---|---|---|
| Wireframe | Low-fidelity layout. No real interactions | Structure and flow |
| Hi-fi design | Polished mockup of the final UI. Still not functional | Look and feel |
| Prototype | Clickable simulation. May not use real data | Usability and the concept |
| Proof of concept | Shows it can be built. Not meant for users | Technical feasibility |
| MVP | Smallest functional version that delivers value | Real behaviour |
| Alpha / beta / GA | A more stable MVP with a limited audience | Usability and performance |
| Product | Scaled, reliable solution (V1, V2, V3...) | Ongoing value |
The PRD defines what. The prototype tests how.
Your job is defining what's worth validating, not doing the design or the coding.
AI shortens the gap between wireframe, PoC, feedback and launch. You can go faster, but don't skip the steps. Each one is a foundation.
Four steps from PRD to prototype
The old version of this lesson had three steps: map the user flows, draw each screen, connect everything. The PRD now comes first, because it's what you feed the AI.
Brain dump first. AI second.
- Start with a raw brain dump of what you already know: the problem, the user, and your assumptions
- Then use AI (ChatGPT or Claude) to organise, reframe and clarify
- Avoid starting with AI prompts. They "prime" your thinking and lead you, instead of following you
- You drive the strategy. AI helps you express it better
The best PMs use AI to reflect their thinking, not replace it.
A PRD is a thinking tool, not paperwork
There's no perfect PRD. Every company is different. The goal is clear communication that draws ideas out of other people's heads: your team's shared brain.
What goes in the PRD
Product Academy PRD template · [DAVE: template link]
Your brain dump goes in. A first-draft PRD comes out.
Here is my brain dump for a product idea: [paste your notes: the problem, the user, what you already know, your assumptions] Act as a senior product manager. Turn my notes into a first-draft PRD with these sections: 1. Problem 2. Market dynamics 3. Customer / user profiles 4. Measure of success 5. Product features & release 6. Design or prototype 7. Risks and regulations Optional: product vision / mission Keep my wording where you can. Do not invent data. Mark anything you are unsure of as an assumption or an open question. Finish with the 5 biggest assumptions I should test first.
[DAVE: confirm prompt, or swap in the one you use live. Confirm exercise timing.]
A Business Analyst agent and a Designer agent
- Customer segmentation
- Pain points
- User journey
- Market size
- Outputs everything as a table
- Writes the prompt for the designer
- Given your brand guidelines and style guide
- Takes the BA's prompt
- Designs the key flows and screens
- Keeps the UI familiar to your users
From my own agent notes. [DAVE: which tool you set these agents up in, e.g. a Claude Project or custom GPT]
Start from the end goal, then list each step
- Who is the target user, and what are they trying to get done?
- Write the end goal first (e.g. "success page"), then work back through every step to get there
- Add the unhappy paths too: what happens when something fails?
Wireframe or hi-fi? Pick what the test needs
For each screen: the pain point it solves, the solution, and how it ties to the company goal.
Connect the screens into one flow
- Arrows linking each screen to the next
- Decision points (e.g. "is the user signed in?")
- The ability to restart the flow
- Assumptions and open questions on the side
- Call out errors and edge cases
The vibe coding prompt
A prompt I used in my own PM interview prep. The BA agent fills it in, then the Designer agent builds from it.
PRODUCT SOLUTION [What you're building, and how it builds on [your company's strengths / moats]] USERS & PAIN POINTS [Target users and their main pain points] KEY FLOWS - One flow per target user - Start with the end goal, then each step to get there - Navigation arrows between steps - Assumptions and questions on the side KEY SCREENS - Start with 4-5 screens, with arrows linking the screens - For each key feature, explain: the pain point, the solution, and how it ties to the company goal - Give the user the ability to restart GUARDRAILS - Familiar UI/UX, following the brand guidelines - Call out error management and edge cases
The tool stack has changed
The old version of this lesson had a long list of tools. Here's what I actually use now.
Prototype in Lovable, finish in code
The prototype isn't thrown away. It becomes the starting point for the real app.
Other notable tools
Tools change every few months. The process doesn't: PRD, user goals, 4-5 screens, stitch it together. Pick one tool and get good at it.
3 types of prototypes you can build
Every AI product sits somewhere on the spectrum, from a single helper to a connected ecosystem.
| Type | What it is | PM context | Examples |
|---|---|---|---|
| AI agents | Small, focused helpers that complete specific user goals or automate repetitive tasks | PMs building new agents or products | Email summariser, research copilot, code generator |
| AI-enhanced apps | Full applications where AI powers key features in a larger product | PMs in existing SaaS companies | Copilots, AI writing and design features |
| AI connected systems | Agents and tools that share data and context for multi-step workflows | Platform PMs in enterprise | Multi-agent systems, AI integrations |
The Deals Agent: one idea, three prototypes
[DAVE: demo link]
[DAVE: demo link]
[DAVE: demo link]
Same PRD. Same user goal. Three different ways to test it.
Put the prototype back into the PRD
- Link the prototype (e.g. your Lovable link) in the PRD's design section, so people can see how it works
- Update the PRD as the prototype changes. It's a living document
- Bring your designer in to connect strategy and design into one flow
From PRD to prototype, in one slide
Build your first 4-5 screen prototype
- Brain dump your idea, then run the starter PRD prompt
- Write the end goal and each step for one target user
- Fill in the vibe coding prompt and build 4-5 screens in Lovable (or your tool of choice)
- Put the link back into your PRD
- Show it to one potential customer and write down what you learnt
Share it: [DAVE: where to share, e.g. Maven community / next session]
Build with us at Product Academy
- Learn, build and grow with other PMs
- More sessions like this one
- No pressure. Have a look if it's useful
Join: productacademy.io/join
Q&A
Test ideas, not perfection.