PM Mapped — AI Toolkit

These guides cover the tools where AI materially changes the work — not every tool. Each one shows you a step-by-step way to work, worked examples with real prompts, a reusable template, and the common pitfalls. Every guide ends with the judgment that stays yours: where AI helps, and where your own thinking is irreplaceable.

Module 1Foundations5 guides

Product Vision Template→

Draft variations and pressure-test your vision with an ego-free critic.

Jobs-to-be-Done→

Surface candidate jobs from raw interview language, then verify.

User Interviews→

Sharpen your guide and speed synthesis — without fabricating quotes.

User Stories and INVEST→

Draft INVEST-shaped stories and split epics into shippable slices.

Acceptance Criteria→

Generate criteria and the edge cases you'd miss — then verify each.

Module 3Discovery & Research4 guides

Advanced User Interviews→

Cluster across sessions and hunt the disconfirming evidence.

Advanced Survey Design→

Catch biased questions and code open-text at scale.

Affinity Mapping→

Seed and re-cluster your map — then do the sense-making yourself.

The Insight Statement Framework→

Tell a real insight from a restatement of the data.

Module 4Delivery & Execution3 guides

Specification Types→

Draft and translate specs across levels — verify against your system.

The PRD→

Assemble the document and find the gaps before the sprint does.

Acceptance Criteria→

Generate criteria and the edge cases you'd miss — then verify each.

Module 5Metrics & Analytics6 guides

The Drop-Off Diagnostic→

Generate the queries and hypotheses — within the discipline.

The PM Analytics Stack→

Map questions to layers; know when 'what' needs a 'why'.

SQL Fundamentals for PMs→

Turn plain-English questions into queries — and verify the number.

Building Dashboards→

Draft tile queries and chart choices; cut to what drives a decision.

Voice of the Customer→

Theme feedback at scale, weighted by severity not just volume.

SCR Storytelling with Data→

Structure findings into a story and catch your own logical leaps.

Module 6Influence & Leadership7 guides

The Executive Briefing Structure→

Restructure answer-first and compress — keep your real confidence.

Writing for Clarity→

A tireless editor for writing you've already thought through.

Presentation Storytelling→

Build the arc and find the weak point in your own case.

Interview Question Types→

Endless practice questions with framework-level critique.

The STAR Method→

Structure true stories and catch the 'we'-not-'I' trap.

The Metric-Drop RCA→

Drill the diagnosis discipline, not a memorised cause.

Portfolio & Personal Brand→

Produce and repurpose case studies — in your own voice.

Where AI changes a PM's work — and where it does not

These twenty-five guides exist because AI is genuinely useful for a specific slice of product work, and close to useless for another. Being clear about which is which is the difference between a real speed-up and a confident-sounding mess you have to unpick later.

AI is strong at first drafts against a known structure. Acceptance criteria, INVEST-shaped user stories, PRD sections, interview guides, survey questions, SQL against a schema you supply. These are tasks with an established shape, where producing a competent version is laborious and reviewing one is fast. The economics favour generation because your cost of checking is far below your cost of writing.

AI is strong at breadth over a corpus you provide. Clustering forty interview transcripts into candidate themes, coding open-text survey responses, listing edge cases you have not considered, generating counter-arguments to your own strategy. Here the value is coverage, not judgment: it surfaces the candidates and you decide which are real.

AI is weak wherever the answer depends on things it cannot see. Your customers' actual behaviour, your organisation's politics, what your engineers already know is fragile, why the last attempt failed. It will produce a fluent answer regardless, and fluency is exactly what makes this dangerous — a wrong prioritisation reads as convincingly as a right one.

The failure mode to watch is fabricated specificity. Ask a model to synthesise interviews and it will sometimes produce a quote nobody said, or a number nobody measured. The synthesis is often directionally right, which is what makes the invented detail easy to miss. Every guide here includes a verification step for this reason: trace each claim back to the source before it leaves your document.

The judgment stays yours. Deciding what to build, what to cut, whose problem matters most, and what you are willing to be wrong about — that is the work. AI can widen the set of options you consider and cut the time to a reviewable draft. It cannot tell you which option is right for your product, and treating its output as a decision rather than an input is the most common way these tools make a PM worse rather than faster.

Each guide follows the same shape: a step-by-step way of working, worked examples with real prompts you can copy, a reusable template, the pitfalls specific to that tool, and a closing note on where your own thinking is irreplaceable. They are written to sit alongside the framework pages in the Atlas, not to replace them — the guide assumes you already know what the tool is for.