From curious beginner to AI Product Manager with a portfolio that gets interviews
For anyone moving into product who wants to own AI features confidently, write PRDs and metrics that hold up in a room full of engineers, and walk into interviews with four real artifacts to show.
30 chapters, 102 lessons
14-day refund on the yearly plan. Real pricing on the plans page, no surprises.
102
lessons, yours to run
Easy to pick up, built to get you the next job
No coding background needed, and you write code with Claude as you go.
Every section below turns into something a hiring manager recognises, not just notes.
Think the way an AI PM thinks, from day one
- You'll learn to split any feature into deterministic versus probabilistic in under a minute, the exact distinction interviewers probe first.
- You'll map a real AI feature to four distinct roles, and articulate in one clean sentence what an AI PM owns that an ML engineer does not.
- You'll start grading real model outputs as SHIP, EDIT, or WRONG on your first day, no coding background needed, just a browser and a judgement call.
Build the fluency that makes you dangerous in any product meeting
- You'll practise prompting and evaluation hands-on so you can hold a credible technical conversation without writing a line of code.
- You'll apply a repeatable framework for judging any AI opportunity, so you're never the person in the room who can't explain whether something is worth building.
- Every exercise runs in plain language through Claude or ChatGPT, the tools you already have open.
Write the PM artefacts nobody else knows how to write
- You'll produce PRDs and success metrics designed for probabilistic features, the craft a hiring manager will pay a premium for because most PMs can't do it.
- You'll practise launch decisions on AI features using evidence and a structured framework, not gut feel.
- You'll write the fallback and error-state logic that classic specs leave blank, the part that separates a safe launch from a rollback.
Leave with a portfolio and interview answers that are already written
- You'll finish four portfolio artefacts built on one continuous case study, ready to walk through in a screen.
- You'll fill an interview bank throughout the course so your answers to 'what do you own?' and 'how did you decide to ship?' are rehearsed, not improvised.
- You'll practise saying out loud what a hiring manager needs to hear, tying every skill back to a real product decision they'd recognise.
Finish this course and you can do all of this, no prior background required:
- You'll be able to judge any AI feature as safe to ship, needs editing, or not ready, and explain your reasoning to an engineering team without a technical background.
- You'll be able to write a PRD for a probabilistic AI feature including success metrics, fallback states, and error logic, the specific artefact AI PM job descriptions keep asking for.
- You'll be able to apply a repeatable opportunity framework to any AI product idea and justify your recommendation with evidence, not opinion.
- You'll be able to articulate in one sentence the boundary between an AI PM and an ML engineer, the question that decides most AI PM screens before they really start.
- You'll be able to walk an interviewer through four portfolio artefacts built on a single continuous case study, showing real product decisions rather than hypothetical answers.
- You'll be able to make and defend a launch decision on an AI feature in a review with engineers, data scientists, and stakeholders, which is the moment most AI PM candidates fall apart.
102 lessons, 30 chapters
Built by practitioners, not influencers
Every agent, skill file, and dataset is drawn from 25+ years and 50+ Oracle ERP implementations across pharma, manufacturing, semiconductor, and distribution.
Kept current with the stack
Claude Code, skills, and MCP move fast. Your licence includes every system update for as long as you're subscribed. Your system gets better, not obsolete.
You've been applying for AI PM roles and the feedback, when you get any, is always some version of 'not enough AI product experience.' You can't get the experience without the job and you can't get the job without the experience. Meanwhile the JDs ask for things like 'experience defining success metrics for probabilistic features' or 'PRDs for ML-driven products' and you're not even sure what those look like in practice. Classic PM courses teach you to write a spec and QA it, which is exactly the wrong mental model for a feature that's right 88% of the time and wrong the rest. So you're stuck: qualified enough to understand the problem, but without the specific artefacts and vocabulary to prove it in an interview room.
What you'll be able to do, module by module
Every module leaves you able to build and run something real. Here is the syllabus, in the order you work through it.
Curriculum
30 chapters · 102 lessons
Orientation (Pre-Week 1)
You can explain the deterministic versus probabilistic split out loud, name what an AI PM owns that an ML engineer does not, and log your first interview bank entry before Week 1 even starts.
0.1 What Is an AI PM, Really?4 items
- 0.1.1: Why the AI PM role emerged and what broke about classic PM🔒
- 0.1.2: AI PM vs. classic PM vs. ML Engineer vs. Data Scientist🔒
- 0.1.3: A real day in the life, hour by hour🔒
- 0.1.4: Why 'I use ChatGPT a lot' is not the same skill🔒
0.2 Meet the Case Study: Northwind Desk Wants AI4 items
- 0.2.1: The CEO's stated ask🔒
- 0.2.2: Company, product, customers, and competitive pressure🔒
- 0.2.3: The stakeholders you'll deal with across the course🔒
- 0.2.4: What 'success' means to each stakeholder, and where they conflict🔒
0.3 The 4-Week Roadmap3 items
- 0.3.1: Skill map: what is added each week and why in that order🔒
- 0.3.2: The 4 portfolio artifacts you'll walk away with🔒
- 0.3.3: How much time to budget per week🔒
0.4 Setting Up Your Toolkit4 items
- 0.4.1: Accounts and access you'll need🔒
- 0.4.2: Verifying your setup with a first real model call🔒
- 0.4.3: A 'hello world' mini-eval: same question 10 times, log the variance🔒
- 0.4.4: Troubleshooting common setup failures🔒
Week 1: Foundations, Any Product, Any Category
You can apply a repeatable framework to judge whether any AI opportunity is worth pursuing, and articulate your reasoning in language a product team and a hiring manager both understand.
1.1 The AI PM's Real Job: Deciding Where Probability Belongs3 items
- 1.1.1: Deterministic vs. probabilistic features🔒
- 1.1.2: The decision loop: problem, capability, risk, value🔒
- 1.1.3: A shipped AI feature traced back to the decision that justified it🔒
1.2 How to Evaluate Any AI Opportunity4 items
- 1.2.1: The 5-question opportunity framework🔒
- 1.2.2: Where to find the ground truth fast🔒
- 1.2.3: Practice pass: a product category you've never used🔒
- 1.2.4: Common trap: starting from the technology🔒
1.3 What Models Can and Cannot Do4 items
- 1.3.1: A working mental model of LLMs for product decisions🔒
- 1.3.2: The capability probe: 30 minutes of structured prompting🔒
- 1.3.3: Hallucination, variance, and drift: the three failure modes🔒
- 1.3.4: Knowing what AI genuinely cannot fix, and saying so early🔒
1.4 Your First Tool: Prompting as a Discovery Instrument3 items
- 1.4.1: Prompting to interrogate feasibility, not just to generate🔒
- 1.4.2: A scrappy prompt prototype in under 15 minutes🔒
- 1.4.3: Iterating the prototype based on what failures reveal🔒
1.5 Case Study Checkpoint: What Should Northwind Build?3 items
- 1.5.1: Running the CEO's chatbot ask through the 1.2 framework🔒
- 1.5.2: Uncovering the actual highest-value wedge in the ticket data🔒
- 1.5.3: Writing the revised opportunity statement you'd defend to the CEO🔒
1.6 Week 1 Portfolio Artifact: AI Opportunity Assessment3 items
- 1.6.1: Turning the 1.2 and 1.3 frameworks into a reusable template🔒
- 1.6.2: Filling it out live for Northwind🔒
- 1.6.3: Checklist: what makes this artifact interview-ready🔒
Week 2: Hands-On Fluency, Applied Regardless of Category
You can run structured prompting and evaluation exercises, grade model outputs with a clear rationale, and hold a credible technical conversation in a product review without a coding background.
2.1 Prompt Engineering for Product Requirements4 items
- 2.1.1: Demo prompt vs. spec-grade prompt🔒
- 2.1.2: Structuring prompts for reliability🔒
- 2.1.3: Handling edge cases and ambiguous inputs🔒
- 2.1.4: Testing a prompt against multiple real-world variants🔒
2.2 Evals: The AI PM's Core Superpower4 items
- 2.2.1: Why 'it looked good when I tried it' kills AI features🔒
- 2.2.2: Building a 25-case eval set by hand from real inputs🔒
- 2.2.3: Grading rubrics a non-engineer can apply consistently🔒
- 2.2.4: Scoring a prototype and reading the failure clusters🔒
2.3 RAG, Tools, and Agents: What to Ask For and When4 items
- 2.3.1: Retrieval, tool use, and agents explained by what each fixes🔒
- 2.3.2: A decision checklist: prompt-only vs. RAG vs. tools vs. agent🔒
- 2.3.3: Prototyping a retrieval-grounded answer flow with no code🔒
- 2.3.4: The questions to ask engineering, and the answers that worry you🔒
2.4 Cost, Latency, and the Unit Economics of AI Features3 items
- 2.4.1: Tokens, context windows, and model tiers in product terms🔒
- 2.4.2: A back-of-envelope monthly cost model🔒
- 2.4.3: Latency budgets: what users tolerate by use case🔒
2.5 Case Study Checkpoint: Northwind's Prototype Meets Its First Eval3 items
- 2.5.1: Scoping the smallest version of the Week 1 wedge🔒
- 2.5.2: Building the prompt prototype and the 25-case eval set🔒
- 2.5.3: The ugly first results, and turning clusters into an iteration plan🔒
2.6 Side Quest: Re-Run the Playbook in a Second Category3 items
- 2.6.1: Picking a second product category to test transferability🔒
- 2.6.2: Rebuilding the prototype and a 10-case eval with minimal changes🔒
- 2.6.3: Reflection: what had to change, what stayed the same🔒
2.7 Week 2 Portfolio Artifact: A Prototype + Eval Pack3 items
- 2.7.1: Packaging the prototype, eval set, and results🔒
- 2.7.2: Writing a one-page readout🔒
- 2.7.3: Checklist: what makes this pack credible to a technical reviewer🔒
Week 3: The PM Craft for AI, the Part Nobody Teaches
You can write a PRD and define success metrics for a probabilistic feature, including fallback states and error logic, the artefacts most PM candidates cannot produce and most hiring managers are actively looking for.
3.1 Writing a PRD for a Probabilistic Feature4 items
- 3.1.1: Why classic PRDs break on AI features🔒
- 3.1.2: Defining acceptable failure: error budgets and fallbacks🔒
- 3.1.3: Writing the Northwind PRD section by section🔒
- 3.1.4: The 'eval as acceptance criteria' pattern🔒
3.2 Metrics for Non-Deterministic Products3 items
- 3.2.1: Why accuracy alone is a trap, and the metric stack that replaces it🔒
- 3.2.2: The one metric that decides launch, and its guardrails🔒
- 3.2.3: Practice: the metrics section for a different category🔒
3.3 Risk, Trust, and Responsible Shipping4 items
- 3.3.1: The AI risk sweep🔒
- 3.3.2: Designing for graceful failure🔒
- 3.3.3: Working with legal and security without stalling the roadmap🔒
- 3.3.4: Documenting risk decisions so nobody is surprised later🔒
3.4 Stakeholder Alignment: Push Back vs. Ship the Chatbot4 items
- 3.4.1: Reading executive pressure: what 'we need AI this quarter' means🔒
- 3.4.2: Disagreeing with the CEO using eval evidence🔒
- 3.4.3: Negotiating engineering capacity with data instead of urgency🔒
- 3.4.4: When to build it their way anyway, and how to timebox the bet🔒
3.5 Case Study Checkpoint: The Roadmap Collides With Reality3 items
- 3.5.1: The twist: legal flags data use, engineering cuts capacity🔒
- 3.5.2: Re-scoping with the 3.1 PRD and the 3.3 risk framework🔒
- 3.5.3: Communicating the revised plan to the CEO and eng lead🔒
3.6 Presenting an AI Product Decision3 items
- 3.6.1: A decision narrative for mixed technical and non-technical rooms🔒
- 3.6.2: Demoing a probabilistic feature honestly: showing failures🔒
- 3.6.3: Handling tough questions live🔒
3.7 Week 3 Portfolio Artifact: A Launch-Ready AI PRD3 items
- 3.7.1: Assembling the full Northwind PRD🔒
- 3.7.2: Including the 3.5 re-scope story as a credibility signal🔒
- 3.7.3: Checklist: what makes a PRD launch-ready🔒
Week 4: Launch, Learn, and Get Hired
You can make and defend a launch decision on an AI feature using structured evidence, walk an interviewer through four real portfolio artefacts, and answer every standard AI PM screen question from a prepared interview bank.
4.1 The Launch Decision: Shipping With Imperfect Evidence3 items
- 4.1.1: Launch tiers: internal, beta, percentage rollout, full🔒
- 4.1.2: The go/no-go review: evals, guardrails, and risk sign-offs together🔒
- 4.1.3: Kill criteria: deciding in advance what makes you roll back🔒
4.2 Case Study Checkpoint: Northwind Goes Live3 items
- 4.2.1: Making the call with a 92% eval pass rate and one open risk🔒
- 4.2.2: Reading the first week of live data🔒
- 4.2.3: Writing the launch retro🔒
4.3 Post-Launch: The Improvement Loop That Never Ends3 items
- 4.3.1: Monitoring AI features in production🔒
- 4.3.2: Turning user feedback and flagged outputs into new eval cases🔒
- 4.3.3: Prompt fixes vs. retrieval fixes vs. model upgrades🔒
4.4 Turning the Case Study into a Portfolio Piece3 items
- 4.4.1: Structuring the write-up: opportunity, evidence, decision, outcome🔒
- 4.4.2: Recording a short walkthrough video: structure and pacing🔒
- 4.4.3: What to leave out for confidentiality🔒
4.5 The AI PM Interview: What They're Actually Testing For4 items
- 4.5.1: The real skills behind typical questions🔒
- 4.5.2: Common formats: product case, technical screen, take-home, behavioral🔒
- 4.5.3: How to present the Northwind case study in an interview🔒
- 4.5.4: Practice round: a brand-new product, cold, framework applied live🔒
4.6 Your 30-Day Post-Course Plan3 items
- 4.6.1: Where to find AI PM roles, and real ones vs. rebadged listings🔒
- 4.6.2: Tailoring your portfolio and resume around the 4 artifacts🔒
- 4.6.3: A week-by-week outreach and application plan🔒
Most applicants can talk about AI in a product context. Very few can produce a PRD for a probabilistic feature, define a success metric that accounts for wrong outputs, and walk through a real launch decision with evidence. This course gives you exactly those capabilities, the ones that put you in a different pile from every other CV the recruiter opens that morning.
From subscribing to running your first artifact
Subscribe and open the orientation module today
You get instant access to the full four-week course including the pre-week orientation. Start with the first lesson, no setup, no installs, just a browser and a blank doc. You'll do your first real output-grading exercise in the first sitting.
Work through one continuous case study, lesson by lesson
Every module builds on Northwind Desk, the same product, the same decisions, the same growing portfolio. You're not watching slide decks. You're writing PRDs, grading model outputs, and filling an interview bank entry after every lesson. Mentor review and daily live sessions keep you moving.
Ship your first real AI PM artefact and add it to your portfolio
By the end of Week 1 you'll have graded real model outputs, mapped the four roles on a live AI feature, and logged your first interview bank entries. By Week 4 you'll have four portfolio artefacts and a practised answer to every question an AI PM screen is likely to throw at you.
Deciding whether to ship Northwind Desk's auto-reply feature
Grade a sample of real outputs
You open 30 support tickets the model has drafted replies for. You grade each one SHIP, EDIT, or WRONG and note the error pattern. Three replies cite a refund policy the product has never offered. You write one sentence describing what a wrong output costs this customer.
Apply the opportunity framework
You map the feature against your Week 1 framework: what task does this replace, how often is the model right, what does a wrong output cost the user, and is there a human fallback. The error rate is acceptable for low-stakes tickets but not for billing disputes. You split the scope.
Write the PRD with probabilistic constraints
You write a PRD that specifies not a single correct output but the conditions under which the feature ships autonomously versus routes to a human agent. You define the success metric as the percentage of auto-sent replies that required no follow-up correction within 48 hours, not accuracy on a test set.
Make and document the launch decision
In the launch review you present your evidence: output sample, error pattern, defined metric, and scope boundary. The engineer says the model is performing well. You explain why 'performing well' on the model side is not the same as safe to ship unsupervised on billing tickets. You recommend a phased rollout, starting with password-reset and plan-change tickets only.
The toolkit you'll build and run in this program
You leave with working assets, not notes.
Real things you can demo in an interview.
A completed PRD for an AI feature on the Northwind Desk case study, ready to talk through in a screen.
A defined success metrics framework for a probabilistic feature, including the fallback and error-state logic.
A structured launch decision document showing how you weighed evidence and made the call.
An interview bank with practised written answers to the AI PM screen questions most candidates stumble on.
A worked opportunity assessment using the repeatable framework from Week 1, applicable to any AI product.
You'll get the most from this if
Classic PMs who want to move into AI product roles and need the specific craft and vocabulary to make that transition credible.
Early-career product people and recent graduates who want their first AI PM job and need a portfolio to prove they can do the work.
Analysts, consultants, or operations professionals who understand products and users but have never held a PM title and want to make the jump into AI product management.
Anyone who has been told they lack AI product experience and wants to fix that with real artefacts, not just reading more articles about AI.
Yearly plan: 14-day keep-the-system guarantee
14-day refund on the yearly plan. Finish the first two modules, and if you haven't run your first live pipeline, take a full refund.
Straight answers before you subscribe
Do I need a coding or technical background to do this course?
No. Every exercise runs in plain language through Claude or ChatGPT, tools you already have. The course is designed from the ground up for people without an engineering background. The orientation lesson is built around the assumption that you have never shipped an AI feature before, and every technical concept is introduced through a plain-language worked example before you're asked to apply it.
Will this work if I'm not already working at a tech company?
Yes. The entire course runs on a single fictional case study, Northwind Desk, so you don't need access to a real product team or internal data. All four portfolio artefacts are built inside that case study. You can do every exercise on a free Claude or ChatGPT account from anywhere.
What if I start and it's not right for me? Is there a refund?
There is a 14-day refund policy. If you work through the course and decide it's not the right fit within the first 14 days, you can request a full refund. The free preview lesson is available before you buy so you can judge the teaching style and level before you commit.
How does the course stay current as AI products keep changing?
The curriculum is updated to reflect changes in the field, and as a student you get access to those updates. The core frameworks, judging probabilistic features, writing PRDs for AI, making evidence-based launch decisions, are designed to be durable across model generations rather than tied to any one tool or API version.
Can I use the portfolio artefacts I build here in real job applications and interviews?
Yes. The artefacts are yours. They are built on a fictional case study specifically so there are no confidentiality issues with showing them to prospective employers. The course is designed with interview use in mind: the interview bank entries you write throughout the course are intended to become your actual spoken answers in screens, not just coursework you file away.