Copied to clipboard!
Module 7 Level 2: Applied AI Design Systems & Skills intermediate 1 hr

Capstone: Build & Audit a Brand System

Set the standard, then refine a UI against it.

What you'll be able to do

  • Build a brand-principles system as your standard
  • Audit a sample UI against that standard with tied, specific findings
  • Produce an exportable artifact that proves the build-and-refine loop

Capstone: Build a Brand System, Then Audit a UI Against It

This is the Level 2 proof. You build a brand-principles system, then audit and refine a sample UI against it, and export the artifact. You run the full art-director loop: set the standard, critique against it, and iterate until the UI is on-brand. AI assists at every step; you own the standard and the final call.

Recommended tool

claude

Handles the whole loop: confirm the standard, run structured critique against it, and write the on-brand revision plan and exportable artifact. Keep your DESIGN.md in the chat so every critique cites a principle or token.

Bring to the exercise

  • Your DESIGN.md from the brand-system exercise (or build a quick one here)
  • A sample UI to audit (a screenshot description, a page you have, or a simple mockup)
  • The three brand principles you will hold the UI to
Step 1

Confirm the Standard

Lock the DESIGN.md as the standard so every later critique cites it, not opinion.

What to substitute before pasting

  • [PASTE YOUR DESIGN.md] Your DESIGN.md from the brand-system exercise, or a quick one built earlier in this session.
Here is my brand system:
[PASTE YOUR DESIGN.md]

This is the STANDARD I am holding a UI to. Confirm you understand it by listing back, in one line each:
- The three brand principles
- The color roles
- The three rules from my Do / Dont you think are easiest to violate

Every critique you give later must cite a principle, a token, or a do/dont rule. No opinions that are not tied to my standard.
What good output looks like
  • Plays back your principles, color roles, and the easiest-to-break rules.
  • Commits to citing your standard on every later critique.

Verify before using AI's output

Step 2

Audit the Sample UI

Critique the UI against the standard, every finding tied to a principle or token.

What to substitute before pasting

  • [DESCRIBE OR PASTE THE SAMPLE UI] The UI you are auditing: a screenshot description, an existing page, or a simple mockup.
Here is the sample UI to audit:
[DESCRIBE OR PASTE THE SAMPLE UI]

Audit it against my standard. For each issue:
- What is wrong (specific element)
- Which principle, token, or do/dont rule it breaks
- Severity (high / medium / low)

Only flag things tied to my standard. If the UI actually honors the standard somewhere, say so, do not invent problems.
What good output looks like
  • Each issue names a specific element and the rule it breaks.
  • Severity tagged high / medium / low.
  • Calls out what is already on-brand, not just problems.

Verify before using AI's output

Step 3

Write the Refinement Plan

Turn the audit into an ordered, on-brand revision plan you approve.

Turn the audit into a REFINEMENT PLAN. For each high and medium issue:
- The exact change to make
- The token or rule it should now follow

Order them highest impact first. Where a fix is a judgment call (which way is more on-brand), present the options and let me decide rather than picking for me.
What good output looks like
  • An ordered, specific change list tied to tokens and rules.
  • Highest-impact fixes first.
  • Judgment calls handed to you as options, not decided silently.

Verify before using AI's output

Step 4

Export the Artifact

Produce one exportable capstone artifact that proves the full loop.

Assemble the capstone artifact in one document:
1. The brand system summary (principles + tokens)
2. The UI audit (issues, rules broken, severity)
3. The refinement plan (ordered changes)
4. A short "what I decided and why" note in my voice on the judgment calls

Clean and exportable. This is my Level 2 proof that I can build a standard and refine a UI against it.
What good output looks like
  • One clean, exportable artifact with all four parts.
  • Includes your own decisions, not just AI output.
  • Reads as proof of the full build-and-refine loop.

Verify before using AI's output