The journal

AI + UX · 8 min read

From messy brief to testable prototype—with AI.

A practical workflow that keeps evidence and judgment in human hands.

AI is most useful when you give it a specific job inside a clear design process. A quick first draft is valuable if it helps you ask better questions, compare options, or test sooner. Treat generated output as material to inspect.

1. Clarify the brief

Provide the user goal, known constraints, and verified evidence. Ask the model to list assumptions, missing information, and competing interpretations. Review that list with your product partner before treating it as a design brief.

2. Explore distinct approaches

Ask for three genuinely different interaction models and the tradeoffs of each. For appointment booking, those might be time-first, provider-first, and guided matching. Judge them against user needs and technical constraints. More options do not automatically mean better options.

3. Make a focused prototype

Use a design or coding assistant to draft one task flow with realistic content. Specify success, empty, loading, and error states. Keep the prototype small enough to inspect. Check every interaction before putting it in front of a participant.

4. Test with real people

Write neutral tasks and observe relevant participants. Generated personas and simulated interviews can help you think of questions; they do not establish what your actual users need or how they behave.

5. Review, document, and improve

Keep a decision log with the evidence, options considered, choice made, and open questions. Check accessibility, content accuracy, and edge cases. Share what you learned with engineering and product management.

A prompt to adapt

Act as a critical UX collaborator. User goal: [goal] Verified evidence: [evidence with source IDs] Constraints: [constraints] Identify assumptions and missing information. Propose three different approaches with tradeoffs. Distinguish supplied evidence from your suggestions. Do not invent research findings. End with a small prototype experiment for the riskiest assumption.

Before sharing project material

Use tools approved for your work. Remove unnecessary personal and confidential details, confirm consent and data handling requirements, and work with synthetic examples where possible. You remain responsible for what you submit and use.

Further reading: NN/g: Using AI for UX Work.

Try it yourself

Use a fictional booking product. Ask AI for three approaches, choose one using explicit criteria, and write the question you would test with a real participant.

Keep learning

Let AI find patterns. Keep the evidence attached.