Learn with Sohaib · PM to AI Builder
From product manager to AI builder.
You already have the hardest skill. Learn the rest, and build the things you used to hand to developers yourself.
The idea
You're already halfway there.
The hardest part of building with AI isn't the code. It's knowing what to build and explaining it well. That has been your job all along.
What you already do
- Understand what the client actually needs
- Turn it into user flows
- Break flows into user stories and specs
- Explain it clearly to the people who build
What AI needs from you
- Full context, with nothing assumed
- A clear flow from the first input to the output
- The work broken into small, testable steps
- One clear explanation at a time
You used to explain things to developers. Now you explain them to AI, and build it yourself.Your explanation becomes the product.
How it works
Five things that change how you build.
- 01
Teach it like a kid
Assume the AI knows nothing. Give it every piece of context first, then ask. The answers get dramatically better.
- 02
Context is the difference
Two people can ask the same AI the same question and get very different results. What changes is the data and context each one brings.
- 03
Know how the data flows
If you can explain how data moves from the first input to the final output, you can build it. This is the core skill of an AI product architect.
- 04
Don't build a wrapper
A thin layer over a model is easy to copy. Your data is the product: the examples, documents and knowledge only you have.
- 05
Keep trying new tools
The tools change every month. The people who keep learning them keep winning, so build in public and learn from what others ship.
- An example
Proposal AI, not a wrapper
Its brain is years of real proposals plus vetted expert advice, transcribed: about 450 MB of text. That context is why it writes like an expert, and why it can't be copied in a weekend.
The method and the tools
Paper-First: from your page to a working product.
The Paper-First Method in three stages, and the tools for each one. The tools will change; the stages won't.
- 01Step 1
Paper
Think before you prompt. One page by hand: the flow, where the AI sits, and where a human must say yes.
- Pen and paperThe flow from first input to output, on one page
- 02Steps 2-4
Pushback
Read the page to the AI and make it disagree, until the plan is honest. Then decide the architecture.
ClaudeYour thought partner, open all day
- 03Steps 5-7
Production
The AI builds inside the structure you decided. You review like an architect, ship, then go back to paper.
LovableDescribe it, get a working app
CursorBuild in the code, with AI beside you- Claude CodeBuild inside your real project
Also in the kit
ChatGPT
GitHub
Supabase
Vercel
n8n
ElevenLabs
Figma
Notion
This week
Your first build.
Take one small requirement you gave developers before, and build it yourself. That's how you see what your specs were missing.
- Step 1
Pick one requirement
One small thing you once wrote up for a developer.
- Step 2
Put it on one page
Draw the flow by hand, from the first input to the output. The free data flow canvas below helps.
- Step 3
Give the AI everything
Paste the page and all the context into Claude, and have it push back until the plan is honest.
- Step 4
Build it
Describe it to Lovable for a working app, or build in the code with Cursor or Claude Code.
- Step 5
Compare
Note everything you had to add that your original spec left out. That gap is what you just learned.
Free for PMs
The Spec to Build Kit, and more.
Paste a spec, get a build-ready brief and your first prompt. Plus the context pack, the data flow canvas and the Paper-First templates. Tell me who you are and I'll send them straight over.
Send them to me
All four, in one email.
FAQ
PMs ask.
Do I need to learn to code?
Not to start. You need to explain clearly and know how the data flows. Start with Lovable, then move into Cursor or Claude Code when you want more control. You'll pick up the code as you go.
What is an AI product architect?
Someone who decides what an AI product should do, designs how data and context flow through it, and can build it with AI. It's the PM skill set, plus the ability to ship it yourself.
I'm a business PM, not a technical one. Is this for me?
Yes. Understanding what the client needs and turning it into a flow is the hard part, and business PMs do that every day. The technical side is the part AI now helps you with most.
Where do I start?
Watch the PM to AI Builder video, get the free templates, and do your first build this week: one small requirement you once gave a developer.