AI for designers
A daily AI workflow that makes your design judgment more useful
Use AI to prepare briefs, explore alternatives, and inspect design details while keeping research, accessibility, and final judgment grounded.
In this guide
- The useful starting point
- Start the day with a decision brief
- Use AI at three useful moments
- A fictional example: appointment rescheduling
- Try a focused design prompt
- Protect the quality of your inputs
- Make the review concrete
- Use the saved time for stronger design work
- When the assistant leads you astray
- Questions you might be asking
- Sources and further reading
The useful starting point
Use AI to clarify the task, generate alternatives, and check a design against explicit criteria. Keep real user evidence separate from generated ideas, inspect the complete interaction, and spend the time you save on better decisions and testing.
Design work includes many tasks that AI can help with: untangling a brief, comparing alternatives, drafting interface copy, and checking whether a flow covers an error state. The value depends on the questions you ask and the judgment you bring to the result.
A practical daily workflow does not need to begin with a new tool. Begin with today's design decision. What is uncertain? What evidence do you have? What must a person be able to do when the design is finished?
Start the day with a decision brief
Write a short description of the audience, situation, task, constraints, and desired outcome. Add what you know from research and what remains an assumption. If the team has an existing design system, include the relevant patterns and content rules.
Ask AI to identify ambiguity before it proposes a solution. A useful response might reveal that “make onboarding simpler” has no defined user, success criterion, or technical constraint.
Keep the brief short enough to use during critique. You are creating a shared reference for decisions, not a document that becomes detached from the work.
Use AI at three useful moments
| Moment | Useful assistance | Your responsibility |
|---|---|---|
| Before designing | Clarify requirements and identify unanswered questions | Confirm the problem and evidence |
| While exploring | Generate alternative flows and wording | Choose based on user needs and constraints |
| Before sharing | Check states, consistency, and unsupported assumptions | Inspect the actual experience and test important behavior |
This keeps AI close to the work without letting it quietly redefine the problem. Generated alternatives are possibilities to inspect, not evidence that users prefer them.
A fictional example: appointment rescheduling
Imagine a designer improving a service website's rescheduling flow. The team has observed that customers sometimes cancel and book again because they cannot find the change option. The designer's task is to make rescheduling understandable while preserving the service's real availability rules.
AI helps list the states: a booking that can be changed, one outside the allowed window, no alternative availability, a failed save, and a successful change. It drafts several explanations for each state.
The designer checks the actual policy and notices that one draft promises a fee waiver the business does not offer. She removes that claim and supplies the approved policy. She then compares two flow options: choosing a new time before confirming the change, or canceling first. The first may preserve more user control, but technical feasibility and actual testing still matter.
The resulting design includes a clear summary of what will change and a way to recover from an unsuccessful save. The improvement came from examining the complete interaction, not merely generating a prettier screen.
Try a focused design prompt
Review this design brief and flow. Separate known evidence from assumptions. Identify missing states, unclear wording, and decisions the user must make. Propose three genuinely different approaches and explain their trade-offs. Use the supplied policy and design-system constraints. Do not invent user preferences, research findings, or business rules. Tell me what needs direct inspection or testing.
Ask the assistant to explain how each option supports the user's task. If all three are the same layout with different decoration, request alternatives in sequence, information hierarchy, or decision structure instead.
Protect the quality of your inputs
Use approved research notes and remove unnecessary personal information. Preserve the source and context of important observations. If you summarize interviews with AI, compare the synthesis with the original notes and check for contradictory examples.
Give the assistant real constraints. A design that assumes an unavailable integration or an impossible policy change may be attractive but unhelpful. Ask it to mark dependencies rather than quietly treating them as solved.
When using images or screenshots as input, inspect details yourself. The assistant may miss small text, state changes, or behavior that is not visible in a static image.
Make the review concrete
- Task completion. Can a person understand what to do and finish the intended job?
- State coverage. Are loading, empty, error, permission, and success states handled where relevant?
- Content accuracy. Do labels, prices, policies, and promises match the source of truth?
- Interaction quality. Can people recover, change their mind, and understand the consequence of an action?
- Accessibility. Inspect keyboard behavior, focus, labels, contrast, and assistive-technology needs using appropriate methods.
- Evidence. Which choices are supported, and which should remain testable hypotheses?
An AI critique can help you prepare this review. It cannot establish that an interface is accessible or usable merely by declaring it so. Test the actual implementation with the right tools and people.
Use the saved time for stronger design work
If drafting variations becomes faster, spend some of that time checking the neglected states, observing a real task, or clarifying a difficult trade-off with engineering. The goal is better work, not simply a larger pile of screens.
Keep a small log of useful instructions and mistakes. Note whether the assistance reduced total effort after review, and whether it improved the clarity of a decision. That record helps you choose where AI belongs in your own process.
When the assistant leads you astray
If it invents research, remove the claim and request a hypothesis label. If it overfits to your preferred solution, ask for the strongest alternative and the conditions under which that alternative would win. If its critique is generic, provide a concrete task and a specific state.
When a suggestion conflicts with observed behavior, return to the evidence. Your role is to connect user needs, business reality, and implementation choices. AI can help you explore and inspect those connections, while you remain accountable for the design decisions.
Questions you might be asking
Can AI replace usability testing?
No. It can help prepare scenarios and inspect likely issues, but generated reactions are not observations of real people using the design.
What should a designer automate first?
Try a bounded preparation task such as state inventory, copy alternatives, or brief clarification. Choose work you can evaluate against explicit criteria.
Can AI certify that my design is accessible?
No. Use appropriate accessibility evaluation methods on the actual interface, including human inspection and relevant user testing. AI feedback can assist but cannot certify the result.
Sources and further reading
- Microsoft human-AI interaction guidelines
- Microsoft appropriate reliance research
- Anthropic evaluation guidance
Sources checked October 16, 2026. Tool features and plan availability can change; review the linked publisher documentation before connecting an account.
Put one idea to work
Take the next step with a practical worksheet, a clear research question, and a standard you can use to review the work.
Explore design worksheets →Explore more from UX Design Coach