AI for Product Managers masterclass

Where AI helps in product work and where it misleads, what never goes into an AI tool, and the habit of writing the bar first.

1. Where AI fits in product work · 12 minThe learner can map the tasks in their own product week to what AI does well, what it does badly and what it must not decide.
2. How AI fails in product work · 14 minThe learner can name five failure patterns of AI in product tasks and say how each would show up in their own work.
3. What never goes into an AI tool · 12 minThe learner can decide which product material may go into which AI tool, and what must be removed or kept out first.
4. The habit behind both halves: write the bar first · 12 minThe learner can write a short quality bar for an AI task before seeing any output, both for their own use of AI and for a product feature.

Discovery synthesis, feedback analysis, specs, prioritisation, competitive analysis, roadmaps, stakeholder updates and experiments, each as a repeatable workflow with checks.

1. Synthesise discovery research without losing the evidence · 15 minThe learner can use AI to synthesise discovery research so that every theme traces back to what customers actually said.
2. Analyse interviews and customer feedback at volume · 15 minThe learner can tag a large batch of customer feedback against a codebook with AI and produce counts they made themselves.
3. Draft a spec or PRD with AI as the writer, not the author · 15 minThe learner can use AI to draft and critique a product spec while keeping every requirement and decision their own.
4. Prioritise with AI as a sparring partner · 12 minThe learner can use AI to test a prioritisation decision by surfacing arguments and assumptions, without letting it score the backlog.
5. Competitive analysis from public sources · 13 minThe learner can produce a competitor comparison with AI in which every claim is sourced, dated and checked.
6. Roadmaps and stakeholder updates · 12 minThe learner can draft roadmap narratives and stakeholder updates with AI that make no promise or date they did not intend.
7. Design experiments and read results with AI · 15 minThe learner can use AI to critique an experiment design and help read the results, while the decision rule and the arithmetic stay checked.

The tool types for product work, choosing a model for a feature, estimating what it costs to run, and justifying a tool choice. Never paid placement.

1. The tool types for product work · 13 minThe learner can name the main types of AI tool for product work, what each is for and where each tends to fail.
2. Choose a model for an AI feature · 14 minThe learner can compare candidate models for a feature on quality, speed, context and cost using their own test set and live prices.
3. Estimate what an AI feature will cost to run · 14 minThe learner can build a monthly cost estimate for an AI feature from request volume, text size and live model prices.
4. Pick and justify a tool for a brief · 12 minThe learner can choose a tool type for a product brief and justify the choice on fit, data terms, checkability and cost.

Chained steps, reusable templates, breaking a feature into jobs, quality bars that hold, risk registers, graded test sets and the ship-or-drop decision.

1. Chain steps with a check between each · 14 minThe learner can break a large product task into AI steps with a check after each, so errors are caught before they spread.
2. Reusable prompt templates for product work · 12 minThe learner can write a prompt template for a recurring product task that builds the checks into the instructions.
3. Decide whether AI should do this job, step by step · 14 minThe learner can break a proposed AI feature into steps and decide, for each, whether it needs a model, a rule, a search or a person.
4. Write an AI feature spec with a quality bar that holds · 15 minThe learner can write an AI feature spec whose quality bar sets separate pass marks for ordinary, hard and risk cases, with a rubric for subjective output.
5. Keep a risk register with owners · 14 minThe learner can keep a risk register for an AI feature in which every risk has an example, a decision, an owner, a test case and a signal to watch.
6. Build a graded test set and grade it consistently · 15 minThe learner can build a test set with a grading guide for each case and show that two graders, or a model grader, reach the same results.
7. Read evaluation results and make the ship-or-drop call · 15 minThe learner can read repeated evaluation runs by case type, compare with a baseline and write a ship, fix or drop decision that cites the evidence.

A team prompt library, review rules, the human in the loop, monitoring after launch and what to do when a feature goes wrong.

1. A team prompt library for product work · 14 minThe learner can set up a shared prompt library for a product team in which every entry has an owner, a version, a test example and a note on when not to use it.
2. Review rules: what needs a person's sign-off · 12 minThe learner can write review rules for their team that say which AI-assisted product outputs need a named person's sign-off before they are used.
3. Design the human in the loop · 14 minThe learner can design where a person reviews an AI feature's output, what they see, what they can do and what the feature may never do without them.
4. Monitor an AI feature after launch · 14 minThe learner can set up monitoring for a live AI feature using user signals, sampled human grading, cost and speed, and re-run the test set when anything changes.
5. When the feature goes wrong · 13 minThe learner can run a response to an AI feature failure in production: contain it, find the cause, add the case to the test set and tell the right people.

Raw customer feedback with AI summaries that invent, drop, merge or miscount themes. Find each error and fix it.

1. Sort the feedback: how the lab works · 10 minThe learner can explain what the Sort the feedback lab tests and the four kinds of error it plants in AI summaries.
2. Sort the feedback: a worked round · 15 minThe learner can work through a full round of raw feedback and an AI summary, find every planted error and produce a corrected summary.

Scope a real AI feature with a written bar, a risk register, a graded test set and a ship-or-drop decision, reviewed against a published ten-item rubric.

1. Capstone: the brief and the rubric · 15 minThe learner can state what the capstone requires and how it is marked, using the published rubric, before starting their own project.
2. Capstone: build, test and write the decision · 20 minThe learner can complete the capstone: write the spec, register and test set, run and grade the tests, and submit a decision memo that meets the published rubric.

How the exam works and what passing it leads to.

1. The exam and the certificate · 8 minThe learner can describe how the masterclass exam works, what it tests and what passing it leads to.