PM AI Partner
A framework for using AI as a thinking partner,
not a writing assistant
Ahmed Khaled
github.com/ahmedkhaledmohamed
The Problem
Platform PMs face a unique challenge
- The truth lives in code, but strategy lives in docs
- Nobody can articulate what makes your team different
- Strategy claims capabilities that don't exist (or miss ones that do)
- Deep codebase understanding takes months — time you don't have
- The same questions get asked repeatedly by stakeholders
The Problem
Traditional approaches don't scale
TRADITIONAL PM
Read Docs → Form Opinions → Write More Docs → Hope
❌ Slow iteration
❌ Context lost between sessions
❌ Tribal knowledge
The Framework
The Core Principle
The Framework
Four Sources of Truth
CODE
Technical
"What CAN happen?"
DOCS
Contextual
"What's EXPECTED?"
DATA
Empirical
"What IS happening?"
JUDGMENT
Human
"What SHOULD happen?"
The Framework
Sources often conflict
💻 ≠ 📄
Code ≠ Docs
"We claim X but code shows Y"
📊 ≠ 📝
Data ≠ Strategy
"Strategy says X matters, metrics show Y"
🧠 ≠ 📊
Judgment ≠ Data
"I think X, but data shows Y"
AI helps you find these conflicts. You decide how to resolve them.
How It Works
Ten Agent Skills
Different AI "modes" for different tasks — install with npx pm-ai-partner@latest
How It Works
Agent Skills — Invoke by Name
| Skill | Command | What It Does |
|---|---|---|
| 🤝 Thought Partner | /pm:thought-partner | Brainstorm, explore options |
| 🔍 Technical Analyst | /pm:technical-analyst | Code → product insights |
| ✍️ Writer | /pm:writer | Briefs, updates, emails |
| 👹 Devil's Advocate | /pm:devil-advocate | Stress-test proposals |
| 🛠️ Builder | /pm:builder | Tools, scripts, prototypes |
| 📊 Data Analyst | /pm:data-analyst | Queries, metrics, insights |
| 📋 Product Brief | /pm:product-brief | Structured PRDs |
| 🎯 Meeting Prep | /pm:meeting-prep | Talking points, Q&A |
| 📢 Stakeholder Update | /pm:stakeholder-update | Status reports |
| 🧭 Strategic Clarity | /pm:strategic-clarity | Team identity, charter |
How It Works
Eight Workflow Patterns
1. Strategic Clarity
New role, unclear team identity → Charter, Value Prop
2. Planning Cycle
Quarterly planning → Roadmap, Priorities
3. Codebase Discovery
Understanding what exists → Capability Audit
4. Tool Building
Need internal tools → Working Prototype
5. Communication Prep
Big meetings → Talk Structure, Q&A Prep
6. PM-Eng Bridge
Translating code ↔ product → Engineering Questions
7. Data Investigation
Metrics analysis → Insights, Hypotheses
8. Doc Restructuring
Confusing docs → Cleaner Structure
Live Demo
Let me show you
Cursor + Claude in action
🔍
Explore Code
Semantic search across repos
📊
Query Data
Analyze metrics from chat
✍️
Write Docs
Grounded in reality
Results
What I produced with this system
4 months building clarity for a platform team
Team now has clarity — crisp answers to "what do you do?" grounded in implementation reality.
Pitfalls
What doesn't work
| ❌ Don't | Why It Fails |
|---|---|
| "Write me a strategy doc" | No direction = generic output |
| Accept first draft | Always needs human refinement |
| Skip codebase grounding | Strategy disconnected from reality |
| Use AI for judgment calls | Can't know your organizational context |
| Long sessions, no commits | Lost context, no rollback |
Key Practice
Git as AI Memory
Your repo is the persistent layer; chat is ephemeral.
✅ Commit after each session
Preserve context
✅ Use branches for exploration
Safe experimentation
✅ Review diffs before sharing
Maintain ownership
✅ Write descriptive messages
Future you will thank you
Get Started
Try it today
npx pm-ai-partner@latest
One command installs 10 skills, 4 commands, and 3 hooks.
Works with Claude Code, Cursor, and Codex.
Questions?
AI helps you explore. You decide.
Ahmed Khaled
github.com/ahmedkhaledmohamed/PM-AI-Partner-Framework