I see AI as a thinking tool, not an output tool. Claude accelerates the parts of my process that are about generating options and exploring patterns. The judgment calls remain mine.
What Claude accelerates
The parts about generating options and exploring patterns.
01 — Writing
02 — Code structure
03 — Accessibility considerations
04 — Testing logic
The judgment calls remain mine
The decisions that need a person who owns the outcome.
- Understanding user needs
- Strategic decisions
- Final design direction
- Quality gates
In practice
This looks like:
01 — Using Claude for ideation and copy iteration when establishing a design system or messaging framework
02 — Drafting pattern libraries and component documentation
03 — Generating test cases and edge cases I might miss
04 — Reviewing designs for accessibility compliance and WCAG standards
05 — Exploring implementation approaches before committing to code
Every output has a human review step
The key is that every output has a human review step. I validate that what AI proposed is correct, complete, and aligned with the brief. Sometimes I reject it entirely and start over. That rigor is what makes AI tools useful rather than just fast.
For a user-focused discipline like UX, this approach feels essential.
The tool amplifies your expertise. It doesn't replace your responsibility.