Module 1

PM Fundamentals Through AI Operations

📖 3 Lessons⏱ 60–90 min 🧪 Lab: AI-Assisted Request Triage✅ Quiz: 7 questions

Learning Objectives

Understand the PM role and core responsibilities
Set up your AI-assisted working environment
Use AI for your first PM task: triaging incoming requests
Understand the difference between AI-assisted and AI-native PM work

Lesson 1.1

What Product Managers Do

Product Management is the practice of identifying customer problems worth solving and coordinating teams to deliver valuable solutions that achieve business outcomes.

A PM's core responsibilities map to five activities:

Activity	Description	Time Spent (typical)
Discovery	Understanding problems and opportunities	20%
Prioritization	Deciding what to build and what not to build	15%
Communication	Aligning teams and stakeholders	30%
Execution	Ensuring delivery happens	25%
Measurement	Evaluating outcomes and learning	10%

💡

Key Insight: Communication and execution take up 55% of a PM's time but are the most automatable through AI. Discovery and prioritization take 35% and benefit most from AI augmentation, not automation.

Lesson 1.2

AI-Assisted vs AI-Native PM Work

Dimension	AI-Assisted	AI-Native
Context	Loaded manually each session	Persistent in structured files
Workflows	Ad-hoc prompts	Reusable, versioned procedures
Quality	Varies by prompt skill	Consistent through templates + QA loops
Memory	Lost between sessions	Preserved in knowledge architecture
Output	Raw AI text needing heavy editing	Structured, reviewed, auditable

This course teaches AI-native PM work. The difference isn't using AI more — it's building systems where AI operates within structured context, reusable workflows, and quality gates.

Lesson 1.3

Setting Up Your Environment

If you completed Module 0, your environment is already set up. This lesson provides the full reference for your workspace structure.

Your PM Workspace Structure

flowscale-pm-os/
├── context/
│   └── company.md        ← Company context (always loaded first)
├── workflows/
├── prompts/              ← Reusable AI prompt files
├── memory/
│   ├── decisions/        ← Decision logs
│   ├── research-gaps/    ← Research gap memos
│   ├── qa-checklists/    ← Quality checklists
│   └── ai-failures/      ← AI failure logs
├── deliverables/         ← Weekly lab outputs
└── team/                 ← Team collaboration files

Loading Context into Your AI Tool

When starting any AI session, always begin with:

Load the company context from context/company.md before responding.
All your work should be grounded in FlowScale's strategic goals and constraints.

⚠️

Why This Matters: Without persistent context, every AI conversation starts from zero. With it, AI outputs are grounded in your company's reality from the first prompt.

Lab 1

AI-Assisted Request Triage

Scenario: It's Monday morning at FlowScale. You have 12 incoming requests from Friday–weekend. You need to triage them before the 10 AM standup.

Input Data — 12 Incoming Requests

1. Enterprise customer Acme Corp wants SSO integration. They're our 3rd largest account. CS says they'll churn without it.
2. Support reports 47 tickets about webhook signature verification failures in the past 30 days.
3. Sales wants a "quick win" feature — export to CSV — to close 3 pending deals.
4. Engineering lead says the API rate limiter is causing false positives for high-volume merchants. Needs fix.
5. CEO asks for an AI-powered workflow recommendation engine after reading a competitor's blog post.
6. Designer wants to redesign the onboarding flow based on user testing insights showing confusion at step 3.
7. Security team flagged that merchant API keys don't rotate automatically. Compliance risk.
8. Marketing wants a public API status page for the website.
9. Customer Success asks for bulk user import to help enterprise onboarding.
10. Data analyst found that users who create their first workflow within 24 hours have 3x higher retention.
11. A Tier-1 merchant reported that settlement reports are showing incorrect currency conversion for AED.
12. Engineering wants 2 sprints of tech debt reduction to improve deployment speed.

Step 1

Write a triage prompt. Create prompts/triage.md:

# Triage Prompt

You are a product manager at FlowScale triaging incoming requests.

Load context from context/company.md first.

For each request, classify:
- **Urgency**: Critical / High / Medium / Low
- **Strategic alignment**: Direct / Partial / None / Contradictory
- **Effort estimate**: Small (<1 sprint) / Medium (1-3 sprints) / Large (>3 sprints) / Unknown
- **Category**: Bug / Security / Feature / Tech Debt / Research
- **Recommended action**: Do Now / Queue for Sprint / Add to Backlog / Decline / Needs Research
- **Reasoning**: 1-2 sentences explaining why

Output as a table sorted by urgency (Critical first), then strategic alignment.

Step 2

Run the triage using your AI tool with the company context loaded.

Step 3

Review the AI output and identify:

1 classification you disagree with — explain why
1 thing the AI missed that a human PM would catch
1 thing the AI caught that you might have missed

Step 4

Create a decision log at memory/decisions/2026-W01-triage.md:

# Triage Decision Log — Week 1

## Date
[Today's date]

## Decisions Made
| Request | Decision | Reasoning |
|---------|----------|-----------|
| #1 Acme SSO | Do Now | Churn risk for top-3 account |
| #7 API key rotation | Do Now | Compliance risk |
| ... | ... | ... |

## AI Quality Notes
- AI correctly identified... but I would have also flagged...
- AI missed the connection between #10 and our activation goal
- AI correctly de-prioritized #5 as "Needs Research" rather than "Do Now"

Quality Criteria

Criteria	Excellent	Good	Needs Work
Prompt quality	Structured, context-loaded, reusable	Mostly structured	Ad-hoc, no context
AI output review	Thoughtful critique with specific examples	Some review	Accepted output as-is
Decision documentation	Complete log with reasoning	Partial log	No documentation

Deliverables

Your triage prompt (prompts/triage.md)
The AI output (full triage table)
Your review notes (disagreements, misses, catches)
Your decision log entry (memory/decisions/2026-W01-triage.md)

How to Verify Completion

Your triage table covers all 12 requests with all 6 classification fields populated
The prompt file loads company context and specifies output format
You have written at least one specific disagreement with the AI output (not just "I agreed with everything")
Your decision log has at least 6 entries with decisions and reasoning
AI quality notes contain at least 2 specific observations (one miss + one catch)

Done when: Another PM could read your decision log and understand exactly what was prioritised this week and why — without asking you.

Quiz

Module 1 — Knowledge Check

0 / 7 answered

Question 1 of 7

According to the PM time allocation model, which activity takes up the most PM time and is also the most automatable?

A Discovery (20%)

B Communication (30%)

C Prioritization (15%)

D Measurement (10%)

Correct. Communication takes 30% of PM time and, along with Execution (25%), forms the 55% that is most automatable through AI. Discovery and Prioritization benefit from AI augmentation, not automation.

Question 2 of 7

What is the primary purpose of loading company context into an AI session before starting work?

A To make the AI response longer and more detailed

B To satisfy a formal process requirement

C To ground AI outputs in your company's reality rather than starting from zero each session

D To prevent the AI from accessing the internet

Correct. Without persistent context, every AI session starts from generic knowledge. Loading company context grounds the AI in your specific strategic goals, constraints, and team reality — producing usable, specific output rather than generic text.

Question 3 of 7

In the FlowScale triage scenario, request #10 states: "Users who create their first workflow within 24 hours have 3x higher retention." Why should this be classified with high strategic alignment?

A Because 3x is a large multiplier and always worth prioritising

B Because the data analyst who found it is trusted

C Because it was submitted first

D Because it directly maps to FlowScale's strategic goal of increasing activation from 38% to 60%

Correct. Strategic alignment means the request connects to a defined company goal. FlowScale's #1 goal is increasing activation. A data insight showing that early workflow creation drives 3x retention is directly actionable for that goal.

Question 4 of 7

What does a PM's decision log serve as, beyond recording what was decided?

A A compliance document for legal review

B Organisational memory — context that prevents re-litigating the same decisions and shows reasoning over time

C A performance review record showing how many decisions were made

D A backup in case the AI tool loses conversation history

Correct. Decision logs are organisational memory. When team members change, strategy shifts, or the same question resurfaces, a good decision log shows what was decided, why, and what the alternatives were. This is a core PM OS component.

Question 5 of 7

Request #5 in the lab is: "CEO asks for an AI-powered workflow recommendation engine after reading a competitor's blog post." What is the most appropriate initial classification?

A Do Now — CEO requests always have highest priority

B Decline — AI features are never worth building

C Needs Research — the request is inspired by a competitor blog post, not by user evidence or strategic analysis

D Add to Backlog — it might be useful in 2 years

Correct. "Needs Research" is the right call. The request is reactive (competitor blog post) rather than evidence-based. Before committing engineering capacity, the PM should research whether this solves a real user problem and whether it maps to FlowScale's goals. HiPPO (Highest Paid Person's Opinion) should not override process.

Question 6 of 7

What is the key characteristic of a well-structured AI triage prompt?

A It is as short as possible to save tokens

B It uses creative, conversational language to get better AI responses

C It asks the AI to think freely and generate its best ideas

D It includes role, context reference, classification categories, output format, and quality constraints

Correct. A structured triage prompt specifies the role ("you are a PM at FlowScale"), loads context, defines exactly how to classify each item, specifies the output format (sorted table), and includes constraints. This produces consistent, reusable output rather than ad-hoc text.

Question 7 of 7

After reviewing your AI triage output, you find the AI correctly classified everything. What should you write in the "AI Quality Notes" section of your decision log?

A "AI was perfect — no notes needed"

B "I did not review the AI output in detail"

C Specific observations about the AI's reasoning, even if the conclusion was right — what assumptions it made, what it might have missed in different circumstances

D Skip the notes section since there's nothing to report

Correct. "Everything was right" is not a useful quality note. Even when the AI is correct, good quality notes describe the reasoning — what assumptions the AI made, which classifications were borderline, and what might go differently with different input data. This builds critical evaluation habits.

Module 1 Score

←

Module 0 — Introduction

→

Module 2 — Product Discovery