Top

Work with Claude like a Great Product Manager

In 2026 a PM who can brief an AI well ships faster and thinks more clearly. Claude can turn a messy meeting into clean issues, draft a PRD, or synthesise ten interviews in minutes. But it only does this well when you give it good context, ask for the right shape of answer, and then check its work with a critical eye. This course is your free, hands-on path to doing exactly that, and then to going one step further: moving from chatting with Claude to delegating work to an agent with a clear task brief.

What you'll be able to do

Prompt with Clarity

Give Claude the goal, the audience, the context and an example, and get useful first drafts instead of generic ones.

Get Structured Output

Ask for tables, user stories with acceptance criteria, or issue-ready text you can paste straight into GitLab.

Review AI Output Critically

Use the 4D habits (Delegation, Description, Discernment, Diligence) to spot gaps, invented facts and weak reasoning.

Run PM Workflows

Draft PRDs, synthesise research, and convert meeting notes into prioritised issues, with you as the editor-in-chief.

Delegate to Agents

Write task briefs an agent can act on, and read what the agent did (its plan, its changes, its summary) without needing to code.

Use AI Responsibly

Know what not to paste in, when to disclose AI help, and how MCP connections let Claude reach your tools safely.

Your certification path ($0, non-technical first)

Every course below is hosted by Anthropic on its learning platform (Skilljar) and shows a free registration button. Certificate details are stated honestly for each step. Total time for the recommended steps 1 to 3 is about 9 hours; plan roughly 12 hours with the practice.

1

Claude 101

Provider: Anthropic Academy · You earn: certificate of completion (listed in the course curriculum) · Cost: Free certificate · Time: about 2 hours (page gives no duration; our estimate).

How to use Claude for everyday work tasks and understand its core features. Start here.

OPEN COURSE
2

AI Fluency: Framework & Foundations

Provider: Anthropic Academy (with Prof. Joseph Feller and Prof. Rick Dakan) · You earn: certificate of completion after the final assessment · Cost: Free certificate · Time: about 4 hours (14 lessons and a quiz, per Anthropic's course listing).

The flagship course. Learn the 4D framework (Delegation, Description, Discernment, Diligence) for working with AI effectively, efficiently, ethically and safely. This is the heart of the PM skill set.

OPEN COURSE
3

AI Capabilities and Limitations

Provider: Anthropic Academy · You earn: course quiz (certificate not confirmed on the course page) · Cost: Free course · Time: about 3.5 hours (13 lessons and a quiz, per Anthropic).

How large language models work and what they can and cannot do. This is what makes your review of AI output sharper.

OPEN COURSE
4

Introduction to Claude Cowork

Provider: Anthropic Academy · You earn: course completion (certificate not confirmed on the course page) · Cost: Free course · Time: about 2 hours (our estimate). Note that the hands-on features may need a paid Claude plan.

Work with Claude on your own files and projects, and learn how to guide multi-step work safely. This is your first taste of delegating to an agent.

OPEN COURSE
5

Optional: Claude Code 101 and Claude Code in Action (read-along)

Provider: Anthropic Academy · You earn: course completion (certificate not confirmed on either course page) · Cost: Free courses · Time: 2 to 4 hours each (our estimate).

These are written for developers and Claude Code 101 expects some comfort with a code editor and command line, so treat them as read-along material to understand what agents do. You do not need to run anything. Skim the lessons so you can read an agent's plan and summary with confidence.

CLAUDE CODE 101 CLAUDE CODE IN ACTION
6

Optional: Introduction to Model Context Protocol (concepts only)

Provider: Anthropic Academy · You earn: final assessment on MCP (certificate not confirmed on the course page) · Cost: Free course · Time: skim the first lessons, about 1 hour.

The course itself is for developers (it builds servers in Python), so do not try the coding. For a PM, the useful idea is this: MCP is how Claude is connected to tools such as GitLab. Read the plain-language intro on the MCP site to understand that.

OPEN COURSE MCP INTRO (PLAIN ENGLISH)

Course catalogues change. If a certificate is not mentioned on a course page, treat it as a free course and screenshot your completion. Catalogue: academy.claude.com (also reachable from anthropic.com/learn).

Practice in GitLab

Do these on your real project board. No code and no new projects needed. Never paste confidential customer data or secrets into a chat.

Exercise 1: Meeting transcript to issues

Take a messy meeting transcript or notes (a real one with sensitive details removed). Ask Claude to: (a) list decisions, (b) list action items with an owner and due date where stated, and (c) draft each action as an issue with a title, a one-paragraph description and 2 to 4 acceptance criteria, in a format you can copy. Tell it to write "unknown" instead of guessing. Create the issues on your board.

Exercise 2: The AI-assisted label

In your project, create a label named AI-assisted (Manage → Labels). Add it to every issue Claude helped draft. Add a second scoped label such as ai-review::pending and change it to ai-review::done only after you have checked the facts, owners and dates yourself. Make a board list for each so you can see what still needs your review.

Exercise 3: A prompt library in the wiki

Open your project wiki and create a page called Prompt Library. Add 4 prompts that worked for you: meeting notes to issues, PRD first draft, interview synthesis, and "review this issue for gaps". For each, record the goal, the prompt text, what context to attach, what a good answer looks like, and what you had to fix. Update the page every week.

Exercise 4: A review checklist in an issue template

Ask your team lead where issue templates are kept, or propose one as an issue. Add a short "AI-assisted review" checklist to your issue descriptions: facts verified, names and dates checked, nothing confidential included, acceptance criteria testable. Tick it before you move the issue out of Open. After two weeks, filter by the AI-assisted label and note how many issues needed edits.

The agentic angle: chatting vs. delegating

When you chat, you and Claude go back and forth and you check every answer as it arrives. When you delegate to an agent, it takes many steps by itself (reading, planning, changing things) and you check the result at the end. That means the quality of your task brief matters far more, because you are not there to steer each step. In the AI Fluency language, this is Delegation (what to hand over) and Description (how clearly you say it), followed by Discernment and Diligence when the work comes back.

A good task brief has six parts

Goal (the outcome, and why) · Context (links to the issue, the doc, the people) · Scope (what is in and, just as important, out) · Constraints (do not touch X, ask before Y) · Definition of done (testable checks) · Report-back (summarise what you did, what you were unsure about, and what you did not do).

Practical exercise: brief, read, review

Pick one small, low-risk issue (for example "tidy the labels on the board" or "summarise last sprint's closed issues"). Write a six-part task brief as an issue comment. If your team has an agent set up, hand it over. Otherwise, ask Claude to act as the agent and reply with the plan it would follow. Then read the plan or the agent's summary critically: Did it stay in scope? Did it say what it was unsure about? What would you add to the brief next time? Save your best brief to the wiki Prompt Library as a template.

Free resource library

BACK TO PILOT PROGRAM BOOK A COACH

Claude Academy FAQ

Do I need to pay for anything?

No. Each course shows a free registration button. Some hands-on features inside Claude itself may need a paid Claude plan, so check before you rely on them. The free courses and the practice exercises can be done on whatever access your team already provides.

Claude 101 and AI Fluency: Framework & Foundations state a certificate of completion on their course pages. For the others the page does not say, so we do not promise one.

Steps 1 to 4 are written for everyone. The Claude Code and MCP courses are for developers, so we ask you only to read along for the ideas.

Passwords, keys, personal data about customers or colleagues, and anything your team has marked confidential. Remove names and details first, and follow your team's AI-use guidance.

Treat it as a fast, capable first draft from a colleague who sometimes makes things up. You stay responsible for the result, so check facts, names and dates before you share it.

Track your progress

Log each step as a to-do in your GitLab workspace: one issue per step, with the label learning::pilot. Add a note with what you learned and a link or screenshot of your completion, then close the issue when you finish the step.

BACK TO PILOT PROGRAM

Your 12-week PILOT path: Weeks 1-2 of 12. Move at your own pace; the order is a suggestion, not a rule.

« PILOT Program overview
Next course: Product Champion »

Ignite Curiosity is a learning community with a focus to make STEAM cool for students of all ages. We share passion, excitement, and a love for learning with our international community.

WHO WE ARE
WHERE WE WORK
  • 125 Frederick St
    Marietta OH 45750
  • 929-229-2917
  • [email protected]
  • Mon - Sat 09:00 - 17:00
A PROGRAM OF

Curiosity Research Corporation is a Non Profit Organization designed to advance education in a P-20+ Initiative.

Copyright 2024 © Ignite Curiosity. Designed by Curiosity Research Corporation.