Module 9 — Building Your PM Operating System

Learning Objectives

Define what a PM Operating System is and why it matters
Identify the five layers of an effective PM OS
Build a personal context file, prompt library, and decision log
Adopt a daily AI workflow rhythm that compounds over time
Avoid the most common PM OS anti-patterns

Lesson 9.1 — What Is a PM Operating System?

A PM Operating System (PM OS) is the personal system of files, habits, prompts, and rituals that enables you to work consistently at a high level — with or without AI. When you add AI, a well-structured PM OS turns a good PM into a force multiplier.

Definition: Your PM OS is the always-on context layer between your brain and your tools. It holds everything an AI agent needs to act on your behalf accurately, and everything you need to onboard, hand off, or resume work instantly.

The Five Layers

Layer	What it contains	Key file(s)
1. Product Context	Product vision, users, problems, strategy	`context.md`
2. Prompt Library	Reusable prompts organised by task type	`prompts/` folder
3. Decision Log	Timestamped decisions + rationale	`decisions.md`
4. Work Templates	PRD, brief, interview, retrospective skeletons	`templates/` folder
5. Rhythm & Rituals	Weekly review, sprint prep, monthly retrospective	`rituals.md`

Compounding returns: Each file you add to your PM OS makes every future AI interaction more accurate. The investment is front-loaded; the payoff is permanent.

Lesson 9.2 — Building Each Layer

Layer 1 — context.md

Your context file is the single most important document in your PM OS. It should answer the questions an AI agent needs before it can help you:

What product are we building and why?
Who are the primary and secondary users?
What problems are we solving today vs. next quarter?
What are the current success metrics and their targets?
What constraints matter most (compliance, tech debt, team capacity)?
What decisions have already been made and are off the table?

Keep it fresh: A stale context file is worse than no context file — it generates confident but wrong outputs. Review and update it every sprint.

Layer 2 — Prompt Library

Store prompts as plain Markdown files named by task. Good naming conventions:

prd-draft.md — full PRD generation prompt
user-story-split.md — epic-to-story decomposition
interview-synthesis.md — convert raw notes to insights
retro-analysis.md — sprint retrospective summariser
competitor-brief.md — competitive analysis template

Each prompt file should include: the task description, the required inputs, the expected output format, and any tone/quality instructions.

Layer 3 — Decision Log

Format each entry as:

## YYYY-MM-DD — [Decision title]
**Decision:** What we decided.
**Rationale:** Why we chose this option.
**Alternatives considered:** What we rejected and why.
**Owner:** Name
**Review date:** When to revisit.

Layer 4 — Templates

Templates are the scaffolding that makes AI outputs consistent. A template tells the AI the exact structure you want every time, so you stop reformatting outputs manually.

Layer 5 — Rhythm & Rituals

Cadence	Ritual	Time box
Daily	Morning brief: review context, top 3 priorities, one AI task	10 min
Weekly	Sprint prep: update context, triage backlog with AI, capture new decisions	30 min
Monthly	PM OS review: prune stale prompts, update templates, log retrospective	60 min
Quarterly	Strategy refresh: re-align context with OKRs, review decision log	2 hrs

Lesson 9.3 — PM OS Anti-Patterns

Anti-pattern	Symptom	Fix
Context overload	context.md is 10,000 words	Split into product-context.md + team-context.md; keep each under 500 words
Prompt hoarding	100 prompts, none labelled, none tested	Keep only prompts you've used in the last 30 days
Decision log neglect	Empty after month 1	Add a sprint-end ritual: log one decision every Friday
No review cadence	Stale files, outdated metrics	Add a calendar block: "PM OS Review" every 4 weeks
Tool dependency	PM OS only works in one specific tool	Keep all files in plain Markdown — portable across any tool

PM OS ≠ productivity system: Your PM OS is specifically about making AI-assisted product work accurate and repeatable. It is not a task manager or a personal knowledge base. Keep them separate.

Lab 9 — Build Your PM OS Starter Kit

You will create the foundational files for your PM OS using the FlowScale context as the working example.

Create your context.md

Open your AI agent and start a new file called context.md. Use this prompt:

Prompt: "I'm a PM for FlowScale, a B2B SaaS platform for subscription billing. Help me create a structured context.md file that covers: product vision, primary users (finance ops teams at mid-market SaaS companies), key problems we solve today (failed payment recovery, dunning management, revenue recognition), current success metrics (recovery rate, churn reduction, MRR expansion), and top constraints (SOC 2 compliance, 3-month roadmap freeze). Keep it under 400 words."

Review the output, edit for accuracy, and save it.

Create your first prompt library entry

Create a file called prompts/interview-synthesis.md with the following structure:

Prompt: "Create a reusable prompt file for synthesising user interview notes. The prompt should: accept raw interview notes as input, extract top 3 pain points, identify any surprising insights, highlight quotes worth sharing with stakeholders, and output a structured summary in Markdown. Add instructions for tone (professional, neutral) and length (under 300 words)."

Log your first decision

Think of one real or hypothetical product decision. Use your AI agent to help you write the decision log entry in the format from Lesson 9.2. The AI should help you articulate the rationale and alternatives even if your thinking was instinctive.

Write your rituals.md

Use this prompt: "Based on my role as a PM at FlowScale, help me write a rituals.md file covering daily, weekly, monthly, and quarterly AI-assisted PM rituals. Each ritual should specify: name, frequency, time box, what AI tool to use, what to input, and what output to expect."

Deliverables

context.md — under 400 words, covering all 5 required sections
prompts/interview-synthesis.md — reusable prompt with input/output spec
decisions.md — at least one entry in the correct format
rituals.md — covering all four cadences

How to Verify Completion

context.md is under 400 words and a colleague unfamiliar with FlowScale could understand the product after reading it
Your interview-synthesis prompt produces a useful structured summary when given 5 raw bullet-point notes as test input
Your decision log entry includes decision, rationale, alternatives considered, owner, and review date
rituals.md lists at least 4 cadences with time boxes and AI tool references
All files are in plain Markdown and would work in any AI tool (not tool-specific syntax)

Done when: You can paste your context.md into a fresh AI chat, ask it to draft a sprint goal for FlowScale, and receive an accurate, contextually grounded response without any follow-up clarification.

Module 9 Quiz

7 questions · click an option to answer · review all before checking your score

1. Which layer of the PM OS most directly improves AI output accuracy?

The context file is the single most impactful layer because it grounds every AI interaction in accurate product reality. Without it, even perfect prompts produce generic outputs.

2. How often should you update your context.md?

A stale context file is worse than none — it produces confident but wrong outputs. Updating every sprint ensures your AI interactions stay accurate as priorities and metrics evolve.

3. What is the recommended maximum length for a context.md file?

Context files work best when focused and scannable. Around 400–500 words per file strikes the balance between completeness and AI token efficiency. Split into product-context.md and team-context.md if it grows.

4. What makes a prompt library entry reusable rather than one-off?

Reusability comes from parameterisation — a prompt that clearly states what goes in and what should come out can be applied to any similar situation. Prompts without this specification produce inconsistent results.

5. Which PM OS anti-pattern describes having 100 prompts, none labelled, none tested?

Prompt hoarding is the accumulation of prompts without curation. The fix is to keep only prompts used in the last 30 days — quality and accessibility beat quantity every time.

6. What is the primary purpose of a decision log?

The decision log's core value is organisational memory — capturing why a decision was made so future you (or a new team member, or an AI agent) can understand the reasoning without having to reconstruct it.

7. Why should PM OS files be kept in plain Markdown rather than tool-specific formats?

Tool dependency is a PM OS anti-pattern. If your system only works in one specific tool and that tool changes, disappears, or becomes unavailable, your entire PM OS breaks. Plain Markdown is universal.

Score: 0 / 7