Module 3 of 6 · Intermediate · 9 min read

UI and UX design

Use AI for user research, flows, wireframes, interface copy and design critique, while keeping real users at the centre.

Key takeaways

1AI speeds up research synthesis, flows, wireframes and microcopy.
2It can't replace talking to real users or testing with them.
3Ask for critique against named heuristics and accessibility standards.
4Treat generated screens as starting points to refine in your design tool.

01Where AI helps

The biggest gains are in the parts designers find slow: synthesis, edge cases and copy.

Turning interview notes into themes, personas and jobs-to-be-done.
Mapping user flows and edge cases (empty states, errors, permissions).
Generating wireframes or first-draft screens from a description.
Writing and tightening interface copy and error messages.
Critiquing screens for usability, hierarchy and accessibility.

02The workflow

AI shortens every step except step 5's real user testing, which stays essential.

Gather real inputs: interview notes, support tickets, analytics, screenshots.
Synthesise them into problems and user goals with AI, then check against the raw notes.
Map the flow, including every edge case.
Generate wireframes or screen drafts, then refine them in your design tool.
Critique with AI against heuristics, then test with real users.
Hand off with clear specs and states for developers.

03Tools to reach for

Multimodal models can critique a screenshot directly, which makes quick reviews easy.

Chat assistants for synthesis, flows, copy and critique of screenshots.
AI features inside design tools for generating layouts and variations.
Prompt-to-UI generators that output code components, for quick interactive prototypes.
Accessibility checkers for contrast and structure.

04Prompt: research synthesis

Keep the model tied to the evidence so it doesn't invent insights.

Example

Below are notes from {{n}} user interviews about {{product area}}. Identify the top five problems users describe. For each: a one-line problem statement, how many interviews mention it, and two direct quotes with the interview number. Then list surprising or contradictory findings. Don't include anything not supported by the notes. <notes>{{notes}}</notes>

05Prompt: flows and edge cases

Edge cases are where products feel unfinished. Ask for them explicitly.

Example

Map the user flow for {{task, e.g. inviting a teammate}} in {{product}}. List each step as screen → action → result. Then list every edge case: empty states, errors, slow networks, permissions, duplicate actions, cancelling halfway, and first-time versus returning users. For each edge case, suggest what the user should see, including the exact message text.

06Prompt: design critique

Attach a screenshot and ask for specific, prioritised feedback.

Example

This screenshot is the {{screen}} of our {{product}}, used by {{users}} to {{goal}}. Critique it against Nielsen's 10 usability heuristics and WCAG contrast and target-size guidance. For each issue: where it is, which principle it breaks, how severe it is (high, medium, low), and a concrete fix. Start with the three changes that would most help a first-time user.

07Pitfalls

AI gives you more options faster. Choosing among them still needs evidence from real people.

Synthetic users: AI-generated personas and 'simulated interviews' are not research. Use them to prepare questions, not to replace users.
Average design: generated screens tend towards generic patterns. Push for your brand and your users' specific needs.
Unverifiable critique: ask for the principle behind each point so you can judge it.
Skipping accessibility: ask for it every time; it is rarely included by default.

Finished reading?

Mark it complete to track your progress, then test yourself with the Build with AI: Software & Media quiz.

Next: Game development →