Welcome to the AI-Native PM Mastercourse

AI is already changing how product managers work. The question is no longer whether to use AI — it's whether you're using it in a way that actually makes you better, or just faster at producing mediocre output.

There's a meaningful difference between AI-assisted PM work and AI-native PM work:

Every module integrates AI workflows directly into PM fundamentals. There is no "traditional PM first, AI later." You will learn discovery, strategy, requirements, and prioritization through AI-assisted execution from Module 1.

The Three-Phase Mastercourse

The AI-Native PM Mastercourse is designed as a complete, progressive learning journey across three phases. Each phase builds on the last. You must complete Phase 1 before moving to Phase 2.

Phase 1 — Foundations: The FlowScale Simulation

All exercises run inside FlowScale, a fictional B2B workflow automation platform at Series B stage, targeting mid-market operations teams. You are assigned as one of three PMs on the team. You will use this company context throughout all 12 modules, building on your work from earlier labs in later ones.

The 13-Module Structure

Weekly Format (Cohort)

• Workshop (60 min): Concepts, frameworks, and guided discussion
• AI Lab (90 min): Hands-on exercise with specific AI prompts and outputs
• Deliverable: One document per module submitted for review

For self-paced learners: work through lessons at your own speed. The lab exercises are designed for solo completion. Allow 3–4 hours per module.

Throughout this course, you will write files with the extension .md. These are Markdown files — plain text files with simple formatting syntax that AI tools, GitHub, and most modern editors can read beautifully.

Markdown matters for AI-native PM work because:

• AI models read and write Markdown natively — it's the best format for structured prompts and context files
• It's version-controlled with Git (so you can track changes to your prompts and context)
• It works in any text editor, on any operating system, forever
• It's faster to write than formatted documents and easier to maintain

Markdown Syntax Cheat Sheet

A Simple Markdown Context File

Here is what a real PM context file looks like in Markdown:

# FlowScale Company Context

## Company
- Name: FlowScale
- Product: B2B workflow automation platform
- Stage: Series B

## Strategic Goals
1. Increase activation rate from 38% to 60%
2. Increase expansion revenue by 25%
3. Reduce onboarding support tickets by 30%

## Current Problems
- 62% onboarding drop-off at step 3
- Weak retention after 30 days

When you paste this file into an AI conversation, the AI understands the structure instantly. This is why we use Markdown instead of Word documents or spreadsheets for context files.

You've probably used ChatGPT or Claude as a chat tool — you type a message, it replies. That's the simplest form of AI interaction. But AI-native PM work goes further, using AI agents.

Chat vs. Agent: What's the Difference?

The Three Components of an Effective AI Agent Setup

How Agents Use Files: The PM OS Model

When you build your PM Operating System (Module 9), you're essentially building a simple agent setup:

Session Start:
  → Load context/company.md       (who is FlowScale)
  → Load context/personas.md      (who are our users)
  → Load context/metrics.md       (what are we measuring)

Task:
  → Load prompts/triage.md        (how should the AI behave)
  → Provide input data            (the 12 requests to triage)

Output:
  → AI produces structured triage table
  → PM reviews, corrects, approves

Memory:
  → Save decision log to memory/decisions/  (persistent learning)

This is AI-native PM work. The AI doesn't need to be re-briefed every session because the context is persistent. The output quality is consistent because the prompt is structured. The decisions are auditable because they're logged.

Tools Used in This Course

These principles apply to every module, every lab, and every AI interaction in this course. Internalise them before you start Module 1.

Principle 1: Ground Everything in Context

AI models have no knowledge of your company, your users, or your constraints. Every useful AI output starts with loading context. A generic prompt produces generic output. A context-loaded prompt produces usable output.

Principle 2: Constrain Outputs

AI will add unrequested content, invent data, and go beyond your scope if you don't constrain it. Every good prompt includes explicit constraints: "Only use the provided data," "Do not invent features," "Output as a markdown table."

Principle 3: Always Review Before Using

AI output is a first draft, not a final answer. Never send an AI-generated document, email, or recommendation without reading it yourself first. AI models can be confidently wrong. Your name is on the output.

Principle 4: Document Failures

When AI produces wrong, biased, or dangerous output, that's a learning signal. Document it in your memory/ai-failures/ log. Over time, patterns emerge — and your prompts improve because of them, not despite them.

Principle 5: Iterate Prompts Systematically

Prompts are code. They have versions. When a prompt produces bad output, you don't throw it away — you improve it and increment the version. prd-v1.md → prd-v2.md. This is how prompt quality compounds over time.

Principle 6: You Own the Output

AI is a tool. You are the PM. The decisions, the quality gates, the ethical judgment — those are yours. AI can draft the PRD. You decide whether it's shipped. AI can score the RICE. You decide whether to trust the score.

Before starting Module 1, complete the following setup. This takes approximately 20–30 minutes.

Step 1: Create Your PM Workspace

Open your terminal (Terminal on Mac/Linux, PowerShell on Windows) and run:

mkdir flowscale-pm-os
cd flowscale-pm-os
git init

Step 2: Create the Folder Structure

mkdir context
mkdir workflows
mkdir prompts
mkdir memory
mkdir memory/decisions
mkdir memory/research-gaps
mkdir memory/qa-checklists
mkdir memory/ai-failures
mkdir deliverables
mkdir team

Step 3: Create Your First Context File

Create a file at context/company.md with this content:

# FlowScale Company Context

## Company
- Name: FlowScale
- Product: B2B workflow automation platform
- Stage: Series B
- Business model: Seat-based SaaS subscription
- ICP: Mid-market operations teams (50–500 employees)
- Primary user: Operations managers

## Strategic Goals
1. Increase activation rate from 38% to 60%
2. Increase expansion revenue by 25%
3. Reduce onboarding support tickets by 30%

## Current Problems
- 62% onboarding drop-off at step 3 (account configuration)
- Weak retention after 30 days
- Enterprise customers demanding advanced permissions
- Increasing competitive pressure

## Team
- 3 PMs, 18 engineers, 3 designers
- 2 data analysts, 6 customer success, 12 sales

## Constraints
- Limited engineering capacity
- Enterprise security requirements
- High integration complexity

Step 4: Test Your AI Tool

Open your AI tool (Claude.ai, ChatGPT, or equivalent). Paste the company context file and then ask:

Based on the company context above, what is FlowScale's most
critical business problem right now, and why?

If the AI answers with reference to the onboarding drop-off, activation rate, or retention — your context loading is working. If it gives a generic answer, check that you pasted the full context file.

Step 5: Verify Git Is Working

git add .
git commit -m "initial: PM OS setup"

If this runs without errors, your version control is set up. You will commit after each lab going forward.

Before moving on, confirm each item below:

• I understand the difference between AI-assisted and AI-native PM work
• I can read and write basic Markdown syntax
• I understand what a system prompt, context file, and AI agent are
• I have memorised (or saved) the 6 AI working principles
• My flowscale-pm-os/ directory exists with all subfolders
• context/company.md is created and populated
• I have tested my AI tool with the company context and received a grounded answer
• Git is initialised and I have made my first commit