Shaping Canvas

Before Implementation

Answer these questions before committing to a full build.

Whose situation should improve?

Stakeholder and user description

What happens today?

Current workflow and pain points

What outcome matters?

Product hypothesis and outcome metric

Why might AI help?

AI-suitability argument

Where should AI not decide?

Deterministic boundaries and human authority

What data and context are required?

Data, grounding and provenance design

What can go wrong?

Failure modes and risk controls

What does good look like?

Acceptance criteria and eval design

What is the smallest end-to-end test?

Prototype or vertical slice

What evidence permits release?

Quality thresholds and accountable decision owner