AI Product Manager.
Prepare product people to discover, build, evaluate, govern and ship AI and ML products - and to use AI to do the PM job itself - with real case studies, an AI product simulation lab and a portfolio project.
8 weeks live class + 4 weeks self-paced.
Two months of live training take you from AI fundamentals to shipping, evaluating and governing real AI products; self-paced weeks turn it into an AI product portfolio through case studies and simulations.
8 live hours a week across 3 sessions, plus self-paced videos, AI case-study deep-dives, the AI product simulation lab and coach feedback. Responsible-AI thinking is woven through every week, not bolted on at the end.
Foundations of AI Product Management
- What an AI product manager does, and how AI products differ from traditional software
- Where AI genuinely creates value - and where it does not
- The AI product lifecycle: problem, data, model, evaluation, deployment, monitoring
- The AI PM’s interfaces: data science, ML engineering, design, legal and leadership
- Probabilistic products: designing for outputs that are right "most of the time"
You’ll leave with a clear picture of the AI PM role and why building with AI is different from building deterministic software - and a mental model for the full AI product lifecycle you’ll use for the rest of the programme.
Spotify Discover Weekly - how a recommendation feature was framed, shipped and measured, and where the PM sat between data science and design.
Leadership wants to "add AI" to your product. Decide whether AI is the right tool for the job, and justify it.
AI & ML Fundamentals for Product Managers
- The ML you must understand as a PM: supervised, unsupervised and generative
- Models, training and inference - the intuition, without the maths
- Why data is the product: quality, quantity, bias and drift
- Classical ML vs deep learning vs large language models - when to use what
- Reading a model card and holding a credible conversation with data scientists
You’ll gain enough technical fluency to make good decisions and earn the trust of your data and ML partners - understanding what models can and cannot do, and why data quality decides everything.
A fintech fraud-detection model - how data quality, false positives and drift shaped the product decisions more than the algorithm did.
Your model is 92% accurate but the business hates it. Diagnose why accuracy is the wrong metric and choose a better one.
The AI-Powered PM: Productivity, Tooling & Prompting
- Using AI copilots to do the PM job faster: research synthesis, PRDs, specs and comms
- Prompt engineering for PM work - patterns, context and iteration
- Prototyping AI features with no-code and design tools to test desirability fast
- Automating discovery: synthesising user interviews, reviews and support tickets
- Where to trust AI, where to verify, and keeping a human in the loop
You’ll leave 10x faster at the parts of the PM job AI does well - and clear-eyed about the parts it does not. This is the "AI productivity" layer competitors like Reforge build whole courses around, applied to your real work.
A PM team that cut research-synthesis time from days to hours with AI - and the guardrails they added to avoid confident nonsense.
Turn 40 raw user-interview transcripts into a validated problem statement using AI, then defend what you did and did not trust.
AI Discovery, Feasibility, Data & the Model Lifecycle
- Framing a problem as an AI problem - and knowing when not to
- Feasibility: is the data available, labelled, representative and legal to use?
- Data strategy: sourcing, labelling, privacy, consent and rights
- Train / validate / test, and why data leakage quietly kills products
- MLOps for PMs: deployment, monitoring, retraining, rollback and cost/latency trade-offs
You’ll be able to run AI-specific discovery, judge feasibility from the data up, and own the model lifecycle decisions - data, deployment, monitoring and cost - that make or break an AI product.
A support-automation product - the feasibility call on whether there was enough clean, rights-cleared data to ship, before a line of code was written.
You have a great AI idea but thin, biased data. Decide: collect, buy, synthesise, or kill it - and make the case.
Building with LLMs, Generative AI & Agents
- LLM product patterns: assistants, copilots, search, RAG and agents
- Retrieval-augmented generation (RAG) and grounding for PMs
- When to prompt, when to fine-tune, and when to build agentic workflows
- Designing for hallucination, latency, cost and safety from day one
- Scoping an LLM or agent feature end to end with engineering and design
You’ll be able to scope and shape generative-AI and agentic features - choosing the right pattern (prompt, RAG, fine-tune, agent) and designing around the failure modes that sink most GenAI products.
A RAG-based support assistant - how grounding, citations and fallback behaviour turned an impressive demo into a shippable product.
Your GenAI feature hallucinates in 5% of cases. Choose the mix of RAG, guardrails and UX that makes it safe to ship.
Evaluation, Metrics & Experimentation for AI
- Offline vs online evaluation - and building an eval set that actually means something
- Quality metrics for AI: precision/recall, groundedness, helpfulness, task success
- Guardrails, red-teaming and regression testing for AI features
- Running A/B tests, shadow launches and staged rollouts for models
- Defining the north-star and health metrics for an AI product
You’ll be able to prove an AI feature is good enough to ship and keep proving it after launch - the evaluation discipline that separates real AI PMs from demo-drivers.
An LLM feature that only shipped once the team built an eval set and guardrails - and how those caught a regression before customers did.
Two models: one scores higher offline, the other wins the A/B test. Decide which ships and explain the gap.
Responsible AI, Ethics, Risk & Governance
- Bias, fairness and representational harm - how to spot and reduce them
- Privacy, security and data-use rights for AI products
- Explainability, transparency and building user trust
- AI safety, risk and the regulatory landscape (EU AI Act and beyond)
- Building an AI risk and governance checklist you can actually use
You’ll be able to ship AI that is fair, safe, compliant and trusted - turning "responsible AI" from a slide into a governance checklist and product decisions, the way leading programmes weave it through every stage.
A biased screening model incident - what went wrong, how it was caught, and the governance that stopped it happening again.
A launch review flags possible bias in your model. Decide whether to delay, mitigate or ship with disclosure - and own it.
AI Product Strategy, GTM, Leadership & Career
- AI product strategy: roadmap, build-vs-buy, and defensibility in the AI era
- Pricing and packaging AI (usage, value and cost-to-serve)
- Positioning and go-to-market for AI products
- Leading AI work: team design, working with research, and org scale
- The AI PM portfolio and interviews: product sense, the AI case, execution and behavioural
You’ll leave able to set AI product strategy, price and take AI to market, lead AI work, and tell your own AI PM story in interviews - closing the loop from idea to offer.
Notion AI - how an AI feature was positioned, priced and rolled out on top of an existing product, and what the PM owned.
Price a new AI feature that costs real money per use. Choose a model that protects margin without killing adoption.
Turn the training into an AI product portfolio.
AI Case Studies & Simulation Lab
- 10 real-shaped AI product case studies
- 10 high-stakes AI product simulations with alternate paths
- Coach feedback on your discovery, evaluation, governance and launch decisions
Apply everything from the live weeks to realistic AI product case studies and the simulation lab - the same situations an AI PM faces on the job.
AI Product Portfolio Project
- Take one AI product from problem to a shippable spec
- Data plan, model approach, evaluation plan, guardrails and go-to-market
- A portfolio piece you can walk any interviewer through
Turn the programme into a portfolio - design, prototype and pitch a complete AI product you scoped, evaluated, governed and can defend, the capstone competitor programmes are built around.
What you walk away with.
Capstone Project · after 8 weeks
Design, prototype and pitch a full AI product: problem, users, data strategy, model approach, evaluation plan, guardrails and go-to-market.
Work Experience Showcase · after 16 weeks
Present the AI product work you delivered inside a real team on Work Experience Showcase Day.
Program Certificate
Earn your UstackSchool AI Product Manager certificate on completion.
Become an AI Product Manager.
Join the next cohort and build the skills, portfolio and proof to ship AI products and get hired.