Context
Get the business context on the page so everything downstream is grounded. Paste the client's URL, use the generated prompt to ask an AI tool (Claude, ChatGPT, Gemini) for a context summary, then paste the response back into the fields below and edit anything that's off.
Where we're heading and how we'll get there
We're not trying to solve everything
One job today. The bare bones first. More comes next round.
Here's how we'll work through it
Seven stations in plain English. The first five explore the problem space; the last two decide what we'll build.
Client snapshot
Fill these in together. The one-line anchor at the bottom is what everything else should ladder up to.
With an API key configured, "Ask Claude directly" fills Summary, Products and Audience for you. Otherwise use Copy prompt and paste the response back.
What do they do, and who do they do it for? 2–4 sentences.
The sentence we'll hold ourselves to for the rest of the session. Finish: "The tool exists so that <who> can <do what> so they can <outcome>."
What we're building this round
Frame what this engagement is going to deliver. Jobs and features in the rest of the workshop should fit within this frame — anything that doesn't is parked for a future round.
What kind of thing is this? Pick the closest match — or "Other" to type a custom type.
What we'll call this tool during the workshop. Doesn't need to be the final brand name — just a handle the room can point at.
Descriptive — what it IS. Different from the anchor sentence above (what it exists FOR).
Regulations & constraints
Capture once, reuse everywhere.
Any laws, regulations or external rules the build has to live inside — Privacy Act, AML/CFT, FSP/FMA, industry standards, client policy. Keep names short and pasteable.
These get tagged against ideas at Stage 5 (so the scoring conversation surfaces them), written into the AI prompt, and show up in every export. Leave empty if the build has no regulatory context — the rest of the tool carries on as normal.
Capture once, reuse everywhere.
Any laws, regulations or external rules the build has to live inside — Privacy Act, AML/CFT, FSP/FMA, industry standards, client policy. Keep names short and pasteable.
These get tagged against ideas at Stage 5 (so the scoring conversation surfaces them), written into the AI prompt, and show up in every export. Leave empty if the build has no regulatory context — the rest of the tool carries on as normal.
List each one as a short name plus (optionally) a note. The short name drives the chip label used later.
Jobs to be done
Get 3–6 "Jobs" the user is trying to do. We're not designing features yet — we're describing the situations someone will be in when they reach for this tool. Format: When [situation], I want to [motivation] so I can [outcome].
Job statements
A job is a situation + motivation + outcome — not a feature.
"Add a dashboard" is a solution. Here's the same thing as a job:
When the end of the month rolls around, I want to see how we're tracking so I can decide what to cut.
Why it matters: if you capture solutions now, every later stage locks them in — the pain conversation becomes "why the dashboard is hard," not "why the month-end decision is hard." Keep the job at the situation level and the MVP shakes out naturally at Stage 5.
A job is a situation + motivation + outcome — not a feature.
"Add a dashboard" is a solution. Here's the same thing as a job:
When the end of the month rolls around, I want to see how we're tracking so I can decide what to cut.Why it matters: if you capture solutions now, every later stage locks them in — the pain conversation becomes "why the dashboard is hard," not "why the month-end decision is hard." Keep the job at the situation level and the MVP shakes out naturally at Stage 5.
Aim for 3–6. If you can't write it as When / I want / So I can, it's probably not a job.
No jobs yet. Click "Add job" to start.
Pain today
For each job, describe how it's handled today, where it breaks, and what the cost is when it breaks. This is how we separate "nice to have" from "actually worth building."
Pain points
One row per pain. The cost column is what sorts MVP from backlog later.
No pain points captured yet. Click "Add pain" to start.
How might we & ideas
Reframe each pain as a "How might we" question, then brainstorm candidate features. Cast wide — we'll sort next. Every idea lands on a card that we'll move into buckets in Stage 5.
How might we…
Open-ended questions, one per pain or job. Good HMWs are neither too narrow ("add a button") nor too broad ("solve money").
No HMW questions yet.
Candidate features
Each card becomes a sortable item in Stage 5. Don't filter yet — if it's been said, write it down.
No features captured yet.
Sort into buckets
Drag each feature — or use the dropdown — into the bucket that fits. The rule: "In the build" is the smallest set of things without which the tool doesn't do its core job.
Unsorted (0)
"In the build" is the smallest set of things without which the tool doesn't do its core job.
If removing an item still leaves a tool that solves the anchor problem, it belongs in "Next version" — not here.
Agree & generate
Sign-off step. Confirm the count, capture the decision, then generate the two outputs — an MVP Scope Statement (.docx) for signature and an Uncosted Backlog (.xlsx).
Sign-off details
Build brief (internal)
Enriches each feature in "In the build" and "Next version" with a user story, scope bounds, complexity signals, acceptance hints, assumptions, dependencies, risks, and regulation touch. Uses one AI call per workshop. Internal artefact — not shared with the client.
Generate outputs
These are written locally in your browser — nothing leaves your machine.
Danger zone
Wipe the current session and start again. Use this when you start a new client.