2.3: Product Strategy

• Time to Complete: 60-90 minutes
• Prerequisites: Module 1.4 (Task Orchestration), Module 1.5 (Custom Sub-Agents)

Start this module in Claude Code: Run /start-2-3 to kick off the interactive experience.

Overview

Module 2.3 teaches you to develop product strategy using AI as a thinking partner. You’ll use Rumelt’s Strategy Kernel framework (Diagnosis → Guiding Policy → Coherent Actions) to structure strategic thinking, pressure-test choices with devil’s advocate, and transform strategy documents into executive presentations.

Key takeaway: AI can’t make strategic decisions for you, but it can help you research faster, apply frameworks rigorously, challenge assumptions, and communicate clearly.

What is Product Strategy?

Product strategy is a set of hard choices about where to focus resources and how to create competitive advantage.

Strategy vs. Not Strategy

AI’s Role in Strategy

Rumelt’s Strategy Kernel Framework

The Three-Part Structure

┌─────────────────────────────────────────┐
│         1. DIAGNOSIS                    │
│    What's the challenge/opportunity?    │
└─────────────────────────────────────────┘
                   ↓
┌─────────────────────────────────────────┐
│       2. GUIDING POLICY                 │
│   What's our overall approach?          │
│   (WHERE to compete, HOW to win)        │
└─────────────────────────────────────────┘
                   ↓
┌─────────────────────────────────────────┐
│      3. COHERENT ACTIONS                │
│   What specific steps will we take?     │
│   (Initiatives that reinforce each      │
│    other and align with the policy)     │
└─────────────────────────────────────────┘

1. Diagnosis

A clear, actionable statement of the strategic challenge or opportunity based on real data.

Good diagnosis:

The AI landscape is heating up fast. Notion, Linear, and Asana have all launched AI features in the past 6 months. As a smaller player with a 2-person AI team and limited budget (~$50k/quarter), we can’t compete on breadth. We need to find a defensible position where we can uniquely win with limited resources.

Bad diagnosis:

Competitors are doing AI (too vague, no insight)

2. Guiding Policy

Your overall approach that makes explicit tradeoffs about where to compete and how to win.

Good guiding policy:

We’ll focus exclusively on voice-first AI for SMBs (5-20 person teams), explicitly NOT competing on breadth against larger players. We’ll differentiate through AI features that small teams need but enterprise-focused competitors won’t build. We’ll subsidize AI costs in our base pricing to drive adoption, accepting lower margins initially to build defensibility through usage data and retention.

Bad guiding policy:

We’ll build great AI features and win in the market (no choices, no tradeoffs)

3. Coherent Actions

Specific, coordinated initiatives that implement your guiding policy and reinforce each other.

Good coherent actions:

Q1 2026:
- Expand voice to meeting notes + task breakdown (deepening voice capability)
- Launch SMB-specific templates (5-20 person team workflows)
- A/B test AI in base price vs. premium tier

Q2 2026:
- Voice collaboration features (multi-user voice sessions)
- SMB admin dashboard (usage tracking for small team leads)
- Refine pricing based on Q1 data

The Strategy Development Workflow

1. COMPETITIVE RESEARCH
   ↓ (Use parallel agents with WebSearch)

2. STRATEGIC CHOICES
   ↓ (Make 5 key tradeoff decisions)

3. DEVIL'S ADVOCATE
   ↓ (Pressure-test each choice)

4. SYNTHESIS
   ↓ (AI creates strategy doc using Rumelt's Kernel)

5. PRESENTATION
   ↓ (Use pptx skill for executive slides)

1. Competitive Research with Parallel Agents

Launch multiple agents in parallel to research different competitors simultaneously.

Spin up 3 agents to research the competitive AI landscape:
- Agent 1: Research Notion's AI features, pricing, and positioning
- Agent 2: Research Linear's AI features, pricing, and positioning
- Agent 3: Research Asana's AI features, pricing, and positioning
 
For each competitor, I need:
- What AI features they've launched
- How they position and price AI
- What seems to be working vs. not working
- Where there might be whitespace for us

Why parallel agents work: 3x faster than sequential research, each agent uses WebSearch for latest information.

2. Making Strategic Choices

Answer 5 multiple-choice questions that force real tradeoffs:

Your 5 answers become your Guiding Policy.

3. Devil’s Advocate - Pressure-Testing Choices

After each choice, use AI to challenge your thinking:

User chooses: A) Go deep on voice
 
Devil's advocate response:
You're betting everything on voice. But what if voice becomes
commoditized in 6 months? OpenAI just released better voice
recognition at 1/10th the cost. If everyone has voice, you've
built your strategy on something that's no longer differentiated.
With a 2-person team, you won't have resources to pivot.
Isn't this risky?
 
User then: Stick with choice OR Reconsider

Why this works: Forces you to think through downsides BEFORE committing. Better to hear hard questions from AI than from your CEO.

4. Synthesizing Strategy Documents

Once you’ve made your 5 strategic choices, AI synthesizes them into a complete strategy document using Rumelt’s Kernel:

AI takes:

• Competitive research → Diagnosis
• Your 5 choices → Guiding Policy
• Creates 6-month roadmap aligned with choices → Coherent Actions

You provide:

• The strategic choices (AI can’t make these)
• Context about constraints and situation
• Judgment about which devil’s advocate challenges matter

Output structure:

# H1 2026 AI Product Strategy
 
## DIAGNOSIS: The Strategic Challenge
[Competitive landscape + your constraints → core challenge]
 
## GUIDING POLICY: Our Strategic Approach
[Your 5 choices synthesized into coherent direction]
[Explicit tradeoffs - what we're NOT doing]
 
## COHERENT ACTIONS: H1 2026 Roadmap
 
### Q1 2026
- Initiative 1 (aligned with choices)
- Initiative 2 (aligned with choices)
 
### Q2 2026
- Initiative 3 (builds on Q1)
 
### Success Metrics
- Metric 1: [Target]
 
## CRITICAL ASSUMPTIONS
- Assumption 1: [What needs to be true]
 
## COMPETITIVE POSITIONING
- Why customers choose us vs. alternatives
- Risks and mitigation

5. Creating Executive Presentations with Skills

Transform your strategy document into presentation slides using the pptx skill:

Use the pptx skill to create an executive presentation from
my strategy document.
 
Requirements:
- 12-15 slides covering complete strategy
- Include: Title, Executive Summary, Diagnosis, Competitive
  Landscape, Strategic Direction, Tradeoffs, Roadmap (Q1/Q2),
  Success Metrics, Critical Assumptions, Why We'll Win, Risks
- Professional design for executive audience
- Save as strategy-review-slides.pptx

Note: The pptx skill requires python-pptx library. If installation fails, you can create a markdown outline of slides instead.

Real-World Example

Scenario: You’re the Gen AI PM at TaskFlow. You shipped AI voice chat for todos (well-received, moderate usage). Leadership asks: How should we evolve our AI strategy for H1 2026?

Approach:

1. Research competitors in parallel:

Spin up 3 agents to research:
- Notion AI (positioning, features, pricing)
- Linear AI (engineering-focused features)
- Asana AI (enterprise focus, strategic direction)

2. Identify diagnosis: Based on competitive research + constraints (2-person AI team, ~$50k/quarter budget):

We can’t compete on breadth with limited resources. Need to find defensible position where we uniquely win.

3. Make strategic choices:

• Focus: Go deep on voice (not breadth)
• Competition: Differentiate for SMBs
• Pricing: Subsidize to drive adoption
• Scope: AI as enhancement
• Risk: Deliberate and defensible

4. Devil’s advocate each choice: After each choice, hear the challenge and refine or stick with it.

5. Synthesize strategy: AI creates complete strategy doc with Diagnosis, Guiding Policy, Coherent Actions.

6. Create slides: Use pptx skill to generate 13-slide executive deck.

Best Practices

Do:

• Base diagnosis on real data (competitive research, user feedback, market trends)
• Make explicit tradeoffs - say what you’re NOT doing
• Use devil’s advocate rigorously to pressure-test every choice
• Ensure coherent actions reinforce each other
• Document assumptions and review quarterly

Don’t:

• Let AI make strategic decisions (that’s your job)
• Confuse goals with strategy (“increase revenue 50%” isn’t strategy)
• Skip hard tradeoffs (real strategy requires saying NO)
• Ignore competitive landscape or set-and-forget strategy

Troubleshooting

My strategy feels generic - could apply to any company

Fix:

• Review guiding policy - does it say what you’re NOT doing?
• Add specific constraints: “With our 2-person team and $50k budget…”
• Test: Could a competitor copy this word-for-word? If yes, it’s too generic.

Devil’s advocate questions feel unfair/overly negative

Remember: Better to hear hard questions from AI than from your CEO. If you can’t defend your choice, maybe reconsider. Use challenges to refine your thinking.

pptx skill won’t install / python-pptx errors

Fix:

1. Try: pip install python-pptx
2. If that fails: pip3 install python-pptx
3. If still failing, create markdown outline of slides instead

My 5 strategic choices contradict each other

Fix:

• Review all 5 choices together - do they tell a coherent story?
• Check for contradictions: Can’t be deep focus AND spread everywhere
• Test coherence: Could you explain all 5 choices in one paragraph that makes sense?

What’s Next?

You now understand:

• Product strategy is about making hard choices (WHERE to compete, HOW to win)
• Rumelt’s Strategy Kernel framework (Diagnosis → Guiding Policy → Coherent Actions)
• How to use AI for competitive research, devil’s advocate, and document synthesis
• How to transform strategy documents into executive presentations with skills

You’ve Completed Module 2 - Practical PM Applications!

Across three modules, you learned the complete AI-powered PM workflow:

How Module 1 Skills Enabled Module 2 Applications:

For the interactive version:

Type: /start-2-3

About This Course

Created by Carl Vellotti. If you have any feedback about this module or the course overall, message me! I’m building a newsletter and community for PM builders, check out The Full Stack PM.

Source Repository: github.com/carlvellotti/claude-code-pm-course