C21. AI Agents Practitioner PM (AI PM Learning Program)
A complete path: lessons, projects, certification, support, and your digital PM portfolio with a free C20 AI Agents Certified PM certificate.
Hey, Paweł here. I’m sharing the second module of our AI PM Learning Program:
AI Agents.
This is where you’ll learn how to design and build LLM workflows, AI agents, and multi-agent systems with AI evals and guardrails.
It’s split into:
C20. AI Agents Certified PM: Demonstrated knowledge and took the exam.
C21. AI Agents Practitioner PM: Demonstrated real experience by building AI systems.
AI Agents Certified/Practitioner PM
Prerequisite: C10. LLM Foundations Certified PM (a test is free to take)
Here’s the exact set of lessons covering the scope you need:
Lesson 1: Introduction to AI Agents for PMs: in this lesson, we cover the basic terms and architectures. Key areas:
What is an agent?
AI agent building blocks: system prompt, LLM, tools/MCP servers, memory, orchestration, UI, evals.
LLM workflows vs. agents vs. multi-agent systems.
Multi-agent architectures: sequential, hierarchical, collaborative, competitive.
Lesson 2: The Ultimate Guide to n8n for PMs: in this lesson, rather than theorizing, we learn the leading no-code framework for building LLM workflows, agents, and multi-agent systems. Key areas:
n8n vs. common agentic frameworks.
n8n nodes overview: trigger, AI, flow, integrations, core, HITL.
Debugging, error handling, special expressions.
Building LLM workflows and AI agents in n8n.
Lesson 3: MCP (Model Context Protocol) for PMs: in this lesson, we explore MCP - often characterized as USB-C for agents. Key areas:
What is MCP (Model Context Protocol)?
MCP server types from this post: remote, stdio (e.g., npx, UV). Also, read about the remote HTTP streamable MCP we haven’t discussed (SSE is now deprecated).
Lesson 4: Building Multi-Agent Systems: in this lesson, we explore how to implement multi-agent systems in practice. We start by reviewing theoretical concepts and example n8n architectures and conclude with a practical exercise. Key areas:
Recommended re-reads (included in the exam):
14 Prompting Techniques Every PM Should Know (focus on agents)
A Guide to Context Engineering for PMs (in particular tools, tool results, memory, RAG, retrieval techniques, assembly techniques with n8n examples)
AI Agent Architectures With n8n Examples - demonstrates implementing some of the theoretical architectures from Lesson 1 in practice.
14 Principles of Building AI Agents - key principles, in particular, the importance of orchestration and choosing the right LLM.
Case Study: Multi-Agent Research System
Lesson 5: A PM’s Guide to Evaluating AI Agents: in this lesson, we explore how to evaluate and implement guardrails for AI agents. Building blocks and key areas:
Coming soon: An extra post about building and evaluating multi-agent systems with OpenAI Agent Builder. The theory is the same, but it will allow you to implement alternative exercises for lessons 3-5. Please note that C20 test requires a basic n8n understanding.
How to Earn the C20 AI Agents Certified PM
Step 1: Earn the C10 AI LLM Foundations Certified PM (prerequisite, free to take)
Step 2: Take a free knowledge test:
20 single-choice questions
80% required to pass
40 minutes
1 attempt per week (a larger pool of questions)
After each attempt you will see your points, mistakes, and explanations so you can learn and improve.
How to Earn the C21 AI Agents Practitioner PM
Step 1: Earn the C20 AI Agents Certified PM (prerequisite)
Step 2: Implement a set of exercises:
Exercise 1: Build a weekly competitor research (A, B, C) from Lesson 2
Tools: n8n, Brave Search API
Exercise 2: Automate Figma → Jira with Claude Desktop from Lesson 3
Tools: Claude Desktop, MCP.
Alternative: Any AI agent with MCP server as a tool, e.g., Cursor or OpenAI Agent Builder.
Exercise 3: How to Build an AI Voice Agent from Lesson 3
Tools: n8n, ElevenLabs
Alternative: Any AI voice agent that interprets your prompts; might be directly in ElevenLabs without the n8n backend
Exercise 4: Build a Multi-Agent Research System from Lesson 4
Tools: Brave Search API, OpenRouter, rapidapi, n8n
Alternative: Any multi-agent system with dynamic agents or sub-agent calls, e.g., using OpenAI Agent Builder.
Exercise 5: AI Evals and Guardrails from Lesson 5
Pick an AI agent you’ve built (e.g., Competitor Analysis, Trello Assistant, AI Voice Agent). Hint: You can use a less capable model to increase the failure rate.
Execute it at least 10 times with diverse inputs (including edge cases like a company that doesn’t exist or a question instead of instruction)
Review traces and document all errors (you won’t reach theoretical saturation, but that’s okay). You can do it by reviewing n8n executions.
Describe each error clearly (wrong output, missing data, bad reasoning, API error, etc.).
Group them into failure modes (e.g., incorrect email format).
Try to eliminate those errors by fixing prompts if instructions were unclear.
Implement one code-based or LLM-as-Judge evaluator.
Implement at least one guardrail.
Submit all five projects together to pawel@productcompass.pm with “C21” in the subject.
Questions?
In case of any questions or doubts, you get full support on our Slack:
#c21-ai-agents-practitioner-pm
Link to join our community: Your Premium PM Resources
I’m also available during weekly office hours.
More About Certificates
After completing each track (e.g., C20, C21), you will get a digital, shareable certificate.
Currently, a student’s public profile looks like this: https://www.accredia.io/users/pawel-huryn
I’m also working on extending accredia.io, so that each certificate can be accompanied by a set of digital assets: your digital PM portfolio. That way, you can demonstrate things you actually implemented beyond taking a quiz.
More Resources & Why Learn With Me
Extra resources
Beyond the lessons, here are some additional resources to explore:
AI Product Manager Glossary - understand key terms
Interactive RAG Simulator - study common RAG architectures visually
AI PM Learning Roadmap - for extra links, guides, and resources (optional)
Why learn with me?
Of course, resources alone aren’t enough - you need support. So why learn with me?
I bring 11+ years of PM experience, including running a successful B2B startup.
But what matters most is practice: rolling up your sleeves is the only way to truly retain knowledge. I’m the only PM author here who spends most of my time actually researching and building with AI - and connecting the two worlds: Product & AI.
I also work closely with Miqdad Jaffer (Product Lead, OpenAI) as an AI Labs Build Leader, and I help AI PM Certification students (#1 AI PM Cohort on Maven) crush their Capstone Projects with demos, live Q&A, and async support.
(30-day money-back guarantee. Many members expense it through their company’s learning budget.)
P.S. Stay tuned, C30. AI-Powered Practitioner PM is next 🙌