AI is changing the speed of product design, but speed alone is not a design strategy. The real opportunity is to use AI to make better questions easier to ask, alternatives cheaper to explore, and evidence easier to synthesize. The designer remains responsible for deciding what matters.
Start with the problem, not the prompt
Before opening an AI tool, define the user, context, desired outcome and constraints. “Make this dashboard better” is too vague to produce useful work. A stronger brief explains who uses the dashboard, what decision they need to make, what information is missing, and what technical or regulatory constraints apply. AI output improves when the design problem is explicit.
Use AI to expand the option space
Ask for several competing approaches rather than one polished answer. Explore different information hierarchies, onboarding strategies, empty states, error recovery patterns and accessibility considerations. Treat the responses as hypotheses. Compare them against user needs, product principles and implementation realities instead of choosing the most attractive suggestion.
Keep evidence in the loop
AI can help summarize interview notes, cluster feedback and suggest patterns, but it can also overstate weak signals or miss context. Keep links to source evidence, inspect representative quotes, and separate what participants actually said from the model’s interpretation. Validate consequential assumptions with users, analytics or experiments.
Make evaluation part of the workflow
Define quality criteria before generating concepts: task success, comprehension, accessibility, trust, time-on-task or business impact. Review multiple alternatives against the same criteria. For interface code or generated prototypes, test responsive states, keyboard interactions, content edge cases and failure paths—not only the happy path.
A practical working rhythm
Frame the problem; generate alternatives; critique them; prototype the strongest options; test with people; then document what changed and why. AI is most useful in the middle of this loop, where it reduces the cost of exploration. It should not erase the beginning—understanding the problem—or the end—validating the solution.
The principle
Design better with AI by increasing the quality of your thinking, not merely the quantity of your output. Let the model help you explore. Let evidence help you decide. Let human accountability guide what ships.