I use AI to reduce the time between an idea and something the team can inspect. It is most useful for exploration, repeated production work, and coded prototypes.
The output still needs design judgement and review.
Give it useful context
I start with the problem, user goal, constraints, and expected output. For interface work, I also provide the component API, tokens, content rules, and accessibility requirements.
I keep sensitive information out of tools that are not approved for it. I use representative test content when realistic data is needed.
For this portfolio, I asked AI to audit every font size in the existing Astro project. The constraint was simple. Keep the visual language while removing page-specific sizes.
Check each step
It first listed the sizes and grouped values doing the same job. We reduced them to 14, 16, 18, 22, and 28px, then mapped them to display, heading, subheading, body, and small text.
The first pass still had page-specific rules. I asked it to trace those selectors, then moved the decisions into shared tokens.
The order matters. Audit first. Agree the model. Change the code. Check every page.
I reviewed the homepage, About, Writing, project, and recipe pages in the browser. I checked hierarchy, wrapping, dark mode, and small screens, then ran the production build.
For product work, I also test keyboard use, focus order, zoom, content wrapping, errors, and different viewport sizes. An engineer reviews code that may move towards production.
AI gets us to a testable version faster. The team still owns the decision and the result.