now on Apple, Spotify, or YouTube.
And check out our awesome sponsors:
-
GibsonAI: Your AI Database Engineer
-
Vanta: Automate compliance, manage risk, & prove trust
-
Maven: You get $100 off Colin’s course by using this exclusive link
I’ve said this many times, and I’ll say it again: “AI won’t replace PMs, but PMs who use AI definitely will.”
And one of the most valuable AI skills you can have in your arsenal is AI prototyping.
That’s why today, we sat down with Colin Mathews — who has taught 8,500+ PMs — to walk through exactly how to build products and features with AI.
What makes this podcast special?
We’re not just talking about it; we’re building it, live, right in front of you.
By the end of this video, I’m CONFIDENT you’ll know exactly how to build products and features with AI and be able to do it yourself.
We go through demoes of ALL 5 major tools out there:
-
0:31 – Bolt Tutorial
-
32:14 – Cursor Tutorial
-
44:29 – Lovable Tutorial
-
58:54 – Replit Tutorial
-
1:11:29 – v0 Tutorial
AI prototyping is making it easier than ever to design, test, and refine features without writing a single line of code.
Whether you’re a technical PM or someone with zero coding experience, AI tools like Bolt AI can help you quickly prototype and iterate on product ideas.
In this guide, I’ll walk through a 10-step process to prototype a paid meeting scheduler, similar to Google Calendar but with built-in payment features, using Bolt AI and screenshots as our input.
Before we dive into AI prototyping, here are 3 key things to keep in mind:
-
AI is powerful, but it’s not magic. The more precise you are about what you want to build, the better the results. Instead of saying, “I want an AI-powered app,” be specific,“I want to create a paid meeting scheduler with an editable details page and a preview feature.”
-
Clarity makes everything smoother. Trying to build everything at once? Bad idea. AI works best when you break things into bite-sized chunks. Start with a core feature, test it, and then layer on additional functionality. This approach prevents AI from getting confused and makes debugging way easier.
-
Finally, AI is here to help you move faster, not to replace strategic thinking. You’re still the decision-maker; AI just makes the process more efficient and creative.
Before building, gather visual references.
Since we want to create a meeting booking app with payment functionality, we’ll start by taking a screenshot of Apollo.io’s meeting scheduler page (or any similar product we want to reference).
This screenshot will serve as a visual guide for Bolt AI.
Now, upload the screenshot into Bolt AI and clearly explain what you want to build.
Example prompt:
“Create a meeting booking app that generates a scheduling link. It should allow admins to edit the details page, and later, we should be able to preview it. Don’t write any code yet — just create a plan.”
Why? Because instead of jumping straight into coding, AI should first generate a PRD (Product Requirements Document), outlining what we’re building and how it will work.
And if you see any problem with the result then ask the AI to correct it, that’s why it is important to ask AI not to code first because it will be harder to undo and redo commands so it’s better this way.
Ask to implement the command after you’re satisfied with the result until then, DON’T CODE!
Bolt AI will generate a Product Requirements Document (PRD) detailing the feature set and implementation plan. This PRD is crucial as it ensures we have a structured roadmap and makes future changes easier.
Pro Tip: Read through the PRD carefully. If something is unclear or missing, refine your prompt and ask Bolt to revise it before moving forward.
Instead of giving Bolt one massive command, break the work into smaller steps. This prevents errors and makes iteration easier.
For example:
-
Phase 1: Build the basic scheduling link feature.
-
Phase 2: Add the admin edit functionality.
-
Phase 3: Implement the preview page.
-
Phase 4: Introduce payment integration.
At each step, review before moving to the next phase.
Now, tell Bolt to align the UI with the Apollo.io screenshot we uploaded earlier.
Example prompt:“Make the interface match the reference screenshot while allowing room for improvements.”
You can tweak design elements before code is written. This makes adjustments easier and avoids unnecessary rework.
We need users to be able to edit their booking settings.
Take a screenshot of the edit feature from Apollo.io, upload it to Bolt, and say:
“Add an edit feature when a user right-clicks on the ‘…’ button.”
Then, take another screenshot of the detailed edit page that appears after clicking on edit button and ask Bolt to implement that as well just like we asked it to add edit feature.
Now, let’s make the meeting booking a paid feature.Again the steps are same almost except the command:
Take a screenshot of a payment page from a similar platform.
Upload it to Bolt and say:
“Include a payment page where users can select a time slot and pay before confirming their booking.”
Make sure Bolt doesn’t write code yet; first, preview the design and functionality.
Before implementing all changes, add a preview section so users can see their booking before finalizing.
Upload a screenshot of a preview section (e.g., from Apollo.io) and tell Bolt:
“Create a preview section that lets users review their details before confirming.”
At this stage, iterate on any final tweaks before deployment.
If something isn’t working as expected, ask Bolt:
“Why didn’t the pricing feature appear?” (if there’s an issue with payments)
“Fix the issue where the edit button isn’t working.”
Or if an error occurs you can also simply ask bolt to solve the issue by explaining what the issue is and how it is intended to work.
Use AI debugging insights to quickly resolve technical problems instead of manually troubleshooting.
Now that your AI prototype is functional, you can track user interactions using product analytics tools like Mixpanel or PostHog.
Two options from here:
→ User Testing: Put it in front of real users and see how they respond.
→Further Iteration: If needed, refine the product using additional screenshots and AI prompts, and you can add as many features as you want we limited it to a few in our step to step guide just to keep things short and simple for you to understand.
Watch everything in action in the podcast by clicking here.
With AI tools like Bolt AI or cursor, prototyping isn’t just for engineers and designers anymore.
Any PM can do it using natural language commands and screenshots.
Instead of waiting weeks to build a prototype, you can iterate in real-time and test ideas faster than ever before.
The future of PM is collaborating with AI to design, test, and refine products in ways we never imagined.
If you prefer to only get newsletter emails, unsubscribe from podcast emails here.
Email productgrowthppp at gmail for advertising opportunities.
I hope you enjoyed the last episode with Brian Balfour (where he explained why AI prototyping is so important). Up next, we have episodes with:
Finally, check out my latest deep dive if you haven’t yet: System Design Interview for (Technical) PMs: How to Ace It
Cheers,
Aakash
