Copied to clipboard!
Module 1 Level 2: Applied AI Design Systems & Skills intermediate 30 min

Building Brand Principles & a Brand System

Your context doc becomes a reusable standard.

What you'll be able to do

  • Build three usable brand principles from your context doc
  • Produce a reusable DESIGN.md with color roles, a type scale, and components
  • Audit the system against do/dont rules so a stranger could stay on-brand

Turn a Context Doc into a Brand System (DESIGN.md)

You have a brand brief or a Level 1 context doc, but every design and every campaign still comes out looking a little different. You need a single source of truth: brand principles plus a reusable DESIGN.md with color roles, a type scale, components, and clear do/dont rules. AI drafts the structure and fills the tokens fast. You stay the art director who decides what the brand actually is and signs off on every choice.

Recommended tool

claude

Strong at turning a loose brief into structured brand principles and a clean, well-labeled DESIGN.md you can reuse. If your reference lives in a long brand PDF, read it in NotebookLM first, then build the system in Claude.

Bring to the exercise

  • Your Level 1 context doc or brand brief (brand name, audience, feeling you want)
  • Any existing colors, fonts, or logos you already use (even rough)
  • Three words you want people to feel when they see the brand
  • One reference brand whose look you admire, and one you want to avoid
Step 1

Set the Rules Before You Paste

Lock the no-invention rule so the system reflects your brand, not a generic one.

I am building a brand system from my own context. I will paste a brief and any existing assets.

Hard rules, confirm before I paste anything:
- Use ONLY what I give you. Invent no brand values, colors, or fonts I did not provide or approve.
- If something a real brand system needs is missing, list it under GAPS. Do not fill it in.
- You draft and structure. I decide what the brand is and approve every choice.
- Plain labels, no jargon I would have to look up.

Confirm these rules in your own words.
What good output looks like
  • Restates the rules in its own words.
  • Commits to parking missing pieces under GAPS instead of inventing them.
  • Acknowledges you make the final brand call.

Verify before using AI's output

Step 2

Draft Three Brand Principles

Turn the feeling you want into three short, usable principles that guide every choice.

What to substitute before pasting

  • [PASTE YOUR LEVEL 1 CONTEXT DOC OR BRAND BRIEF] The Level 1 context doc or brief that describes your brand, audience, and goals.
  • [THREE WORDS] The three words you want people to feel when they see the brand.
  • [REFERENCE BRAND] One brand whose look or feel you admire.
  • [ANTI-REFERENCE] One brand you do NOT want to look like.
Here is my brief / context doc:
[PASTE YOUR LEVEL 1 CONTEXT DOC OR BRAND BRIEF]

The three words I want people to feel: [THREE WORDS]
A brand I admire: [REFERENCE BRAND]. A brand I want to avoid looking like: [ANTI-REFERENCE].

Draft exactly three BRAND PRINCIPLES. Each is one short sentence I could actually use to settle a design argument (for example "Clarity over cleverness"). Under each, write one line on what it means in practice.

Use only what I gave you. If you need something I did not provide, list it under GAPS.
What good output looks like
  • Exactly three principles, each a short and usable sentence.
  • Each principle has a one-line "in practice" note.
  • Principles trace back to your three words, not generic brand-speak.

Verify before using AI's output

Step 3

Build the Design Tokens

Turn the principles into color roles and a type scale you can reuse everywhere.

Using the three principles, draft the design tokens.

COLOR ROLES (not just hex codes, name the job each color does):
- Primary, secondary, accent, surface/background, text, and one state color (success or warning).
- For each: a plain-language role, a candidate hex, and where it should and should not be used.

TYPE SCALE:
- A heading font and a body font (pair them; say why they fit the principles).
- A simple scale: display, heading, subhead, body, caption, with sizes.

Use any colors or fonts I already gave you. Mark anything you proposed as PROPOSED so I know to approve it.
What good output looks like
  • Colors are named by role, not just listed as hex codes.
  • Each color says where to use it and where not to.
  • A font pairing with a reason tied to the principles, plus a clean type scale.
  • Anything AI proposed is marked PROPOSED.

Verify before using AI's output

Step 4

Define Components and Do / Dont Rules

Capture the reusable pieces and the guardrails that keep the brand consistent.

Now draft the COMPONENTS section and the rules.

COMPONENTS: buttons (primary/secondary), inputs, cards, and one nav pattern. For each, describe the look in plain words tied to the tokens (color role, spacing feel, corner style).

DO / DONT: write 5 DOs and 5 DONTs that protect the brand (for example "DO keep one accent color per screen" / "DONT stretch or recolor the logo").

Keep it specific enough that someone who has never seen the brand could stay on-brand from this alone.
What good output looks like
  • Core components described in plain language tied to the tokens.
  • Five concrete DOs and five concrete DONTs.
  • A stranger could stay on-brand using only this.

Verify before using AI's output

Step 5

Assemble the DESIGN.md

Produce one clean, reusable file you paste at the top of every design chat.

Assemble everything into a single DESIGN.md with these sections in order:
1. Brand Principles (the three)
2. Color Roles
3. Typography (pairing + scale)
4. Components
5. Do / Dont

Plain headings, clean formatting, no filler. This is the file I will paste at the top of every future design and campaign chat, so it must be self-contained and skimmable.

End with a short GAPS list of anything still unapproved or missing.
What good output looks like
  • One self-contained DESIGN.md with all five sections in order.
  • Skimmable and clean, no filler.
  • A GAPS list flags anything still open.

Verify before using AI's output