A better brief.
A better thinking partner.
Four complete workflows with inputs, copyable prompts, example outputs and checks that keep judgment with you.
These are tool-agnostic workflows, not a connected AI service. Copy a prompt into a tool you are authorized to use. Remove identifying or confidential material and check its data settings first. Examples are fictional.
Plan research that answers a decision
Bring these inputs
A decision brief, known evidence, intended participants, access needs and recruitment constraints.
A useful output
A neutral interview or task guide plus a list of unanswered questions.
Illustrative example
For a delivery-options study, replace “Would our new delivery calendar help?” with “Tell me about the last time you needed an order by a particular date. How did you work out whether it would arrive?”
Before you use the output
- Read each question aloud and remove the solution from the wording.
- Check that participant criteria match the decision, including relevant access needs.
- Confirm consent, note handling and recruitment before contacting people.
- Run a pilot and adjust the guide based on what it actually elicits.
Turn a generated screen into a useful critique
Bring these inputs
Screens you may share, one user task, constraints, and the intended behavior.
A useful output
A review grouped by observed issues, hypotheses, missing states and questions.
Illustrative example
Observed: a button says Continue. Hypothesis: a buyer may not know whether it charges their card. Next check: ask a participant what they expect, and inspect the actual payment behavior.
Before you use the output
- Open the original screen and verify every claimed observation.
- Check interaction behavior directly; a screenshot cannot reveal focus order or screen-reader output.
- Keep disagreements and tradeoffs visible rather than accepting a single “best” design.
- Use the Design Lab to compare alternatives, then test the important assumption.
Build one working flow with AI
Bring these inputs
A bounded flow, a simple state map, test data and explicit exclusions.
A useful output
A runnable prototype with realistic state transitions and a list of simulated behavior.
Illustrative example
For a booking prototype, choose a slot, review it, simulate an unavailable-slot response, select another slot and show a confirmation. Mark the confirmation “Demo only—no appointment booked.”
Before you use the output
- Run it and follow every link and control; read code you intend to reuse.
- Try long text, invalid input, retry and a narrow viewport.
- Inspect network activity to confirm the prototype uses only approved data and services.
- Test with people before describing the design as validated; production needs separate security and engineering review.
Make the unglamorous states explicit
Bring these inputs
An approved flow, component references, decisions and unresolved engineering constraints.
A useful output
A state-and-behavior inventory with acceptance criteria and open questions.
Illustrative example
Given a failed save, the entered project name remains visible, an error identifies the failure, and Retry is available. Whether retries are idempotent must be confirmed with engineering.
Before you use the output
- Review states with engineering before treating the checklist as agreed scope.
- Replace subjective criteria such as “looks clean” with observable behavior.
- Trace each acceptance criterion to an actual decision or documented requirement.
- Assign owners to unanswered questions and retest the implemented flow.