The Ultimate AI PM Learning Roadmap

An extended edition with dozens of AI PM resources: definitions, courses, guides, reports, tools, and step-by-step tutorials. Updated: Nov 4, 2025.

May 28, 2025

Hey, welcome to the free edition of The Product Compass newsletter.

Every week, I share actionable tips, templates, resources, and insights for PMs.

Consider subscribing or upgrading your account for the full experience:

Updated on: Nov 4, 2025

I refreshed linked resources.

In this issue I cover everything you need to know as an AI Product Manager.

It aggregates resources, templates, and step-by-step guides for AI PMs and AI builders.

If I had to learn AI Product Management again, I would start here:

1. Basic Concepts

Start with understanding what an AI Product Manager is.

Next, for most PMs, it makes no sense to dive deep into statistics, Python, or loss functions. Instead, you can find the most important concepts here: Introduction to AI Product Management: Neural Networks, Transformers, and LLMs.

[Optional] If you want to dive deeper, I recommend you check out an interactive LLM visualization:

[Optional] Finally, as an AI PM you will most likely work with LLMs, as they are the most cost-effective. But just in case, here are 8 other terms you might come across, explained by Generative AI:

  1. LLM (Large Language Models): Great for natural language understanding and generation (think ChatGPT).

  2. LCM (Latent Concept Models): Powerful in capturing nuanced concepts hidden in data.

  3. LAM (Language Action Models): Designed to not just understand, but also take action based on language input.

  4. MoE (Mixture of Experts): Smartly combines expertise from multiple specialized models for superior performance.

  5. VLM (Vision-Language Models): Handles text AND images, bridging visuals and language seamlessly.

  6. SLM (Small Language Models): Ideal for efficiency and speed, especially in resource-constrained environments.

  7. MLM (Masked Language Models): Masters context, great at predicting masked or missing content in text.

  8. SAM (Segment Anything Models): Perfect for precise image segmentation and detailed visual understanding.

2. Prompt Engineering

52% of U.S. adults use LLMs. But very few know how to write good prompts.

I recommend starting with resources curated specifically for PMs:

[Optional] Other generic, free resources:

Share

3. Fine-Tuning

Use those platforms to experiment with training and validation data sets and parameters such as epochs. No coding:

You can practice fine tuning by following this practical, step-by-step guide:
The Ultimate Guide to Fine-Tuning for PMs

4. RAG (Retrieval-Augmented Generation)

RAG, by definition, requires a data source plus an LLM. And there are dozens of possible architectures.

So, rather than studying artificial names, I recommend the following resources to learn RAG in practice:

5. AI Agents & Agentic Workflows

AI agents are the topic you can learn best by doing. I see too many BS advice by people who have never built anything.

My favorite tool, by far, is n8n, that allows you to:

  • Create complex agentic workflows and multi-agent systems with a drag-and-drop interface.

  • Easily integrate with dozens of systems (Google, Intercom, Jira, SQL, Notion, etc.).

  • Create and orchestrate AI agents that can use tools and connect to any MCP server.

You can start with those guides:

[Optional] And here are my favorite, free generic guides and reports:

Refer a friend

6. AI Prototyping & AI Building

I listed many tools, but in practice, Lovable, Supabase, GitHub, and Netlify are 80% of what you need. You can add Stripe. No coding.

Here are four practical tutorials:

[Optional] If you want to build and monetize your products, e.g., for your AI PM portfolio:

When building, focus on the value, not hype. Customers couldn’t care less about whether your product uses or has been built with AI.

7. Foundational Models

My favorite models (August 15, 2025):

  • GPT-5 > GPT-4.1 > GPT-4.1-mini for AI Agents

    • GPT-5 might sound robotic and is slower, but it excels at planning, more reliable tool use, and complex tasks.

  • Claude Sonnet 4.5 for coding

  • Gemini 2.5 Pro for everything else

8. AI Evaluation Systems

You might have the most advanced architecture. But the real question is this: Does your product actually work?

Evals are the most critical element. And it's a task also for PMs.

9. Other Resources

AI Product Strategy & Product Leadership:

A few other AI PM resources that I found particularly useful over the last months:

Share

10. Recommended AI Product Management Courses and Programs

I. AI PM Learning Program

I’m running a hands-on, async AI PM Learning Program focused on learning by doing. You get my full support, a dedicated Slack channel, and weekly office hours.

AI PM Learning Program

The AI PM Learning program is free for paid subscribers of this newsletter.

Learn more, start the first module, and get certified here:

You can continue with:

II. AI Product Management Certification

Highly Recommended: The AI Product Management Certification — a 6-week, hands-on cohort led by Miqdad Jaffer (Product Lead @ OpenAI).

I joined the Spring 2024 cohort and loved the mix of practical projects and networking. Since then, I’ve become an AI Build Labs Leader.

The next session starts January 26, 2026 — and I managed to secure a $500 discount for our community:

Claim a $500 discount

III. AI Evals For Engineers & PMs

I’ve participated in the first cohort together with 700+ AI engineers and PMs. I have no doubt that every AI PM must understand evals in depth. And I agree with Teresa Torres:

The last cohort started on October 10, 2025. I will update the link once there is a new enrollment available.

Learn more

Visual Summary

Thanks for Reading The Product Compass Newsletter

Hope that helps!

It’s great to explore, learn, and grow together.

Have a great rest of the week ahead,

Paweł

Discussion about this post

No posts

Ready for more?