KIMBERLY FLOWERSPRODUCT CREATIVE DIRECTION
Back to selected work

AI ENABLED CREATIVE PRACTICE

Explore widely.
Review deliberately.

Brightpath provides a concrete example of an AI-assisted design loop: a reference, a set of visual decisions, generated assets, critique, and a responsive implementation.

Creative direction and feedback: Kimberly Flowers

ACTUAL DESIGN ITERATION

From expressive studio
to a clearer introduction.

Earlier Brightpath design with a serif headline and abstract ribbon image
Earlier directionArchived design for comparison only. This is not the current Brightpath site.
Current Brightpath with the eggplant-and-rose palette, team-first hero and lightbulb logo
Current directionPhoto-first layout, compact sans-serif hierarchy, rose accents and the selected lightbulb mark.
Open the current Brightpath site

The review loop, made concrete.

  1. Set the frame. Keep the studio offering and use the supplied reference for a photo-first hierarchy.
  2. Constrain the visual language. Retain eggplant; replace lavender with rose.
  3. Generate and compare. Explore alternate marks and select the lightbulb direction.
  4. Reject weak output. An initial transparent icon extraction had edge artifacts and was not published.
  5. Check the actual interface. Correct the mobile hero gap and the Careers image crop in a browser.
  6. Review coherence. Replace repeated photos with distinct scenes and verify each content photograph is used once.

What AI accelerated—and what still needed judgment.

Useful for exploration

  • Generating and refining a controlled asset palette.
  • Comparing logo directions before choosing one.
  • Translating a reference into working responsive layouts.
  • Documenting decisions and preparing alternative interface states.

Human review remained essential

  • Selecting an appropriate direction for the company.
  • Catching repeated imagery and mobile cropping problems.
  • Rejecting visual artifacts rather than calling a generation finished.
  • Checking real behavior and avoiding unsupported outcome claims.

A repeatable practice, not a promised metric.

For a team rollout, I would start with a bounded pilot, an approved source and data policy, review criteria, examples of accepted and rejected work, and a clear owner. Measure useful output, rework, review time and adoption before claiming a productivity gain.

No measured organization-wide adoption, time saving, or production customer outcome is claimed for this portfolio’s AI workflow.