Questions before assets; positioning + de-souping; systems not assets; tone specs.
Graphic design · UX · motion — 4 hours, drop-in friendly
Graphic Design and Branding: Systems, Layouts, Copy, and Production
Use AI to accelerate brand systems, copy variants, and production specs — while understanding precisely why design judgment is the thing clients actually pay for.
01This Week at a Glance
You will make: a mini brand sheet — positioning, voice, type/colour rationale, asset/export checklist — for a micro-brand you care about.
Stress tests; production acceleration; the contrast trap.
Toy Cathedral: a weak "logo for my band" idea becomes a campaign kit.
Your micro-brand system; de-souped lines and ✏️ counts.
The economics of this week
A model produces a positioning line, palette, and three taglines in forty seconds — all competent, all interchangeable. Competent-and-interchangeable is the definition of brand failure. Clients never bought assets; they bought judgment. AI made assets nearly free, which makes judgment the entire remaining product.
02Key Concepts
1 · Questions before assets
The amateur move: "design me a logo." The professional move: interrogate first — audience, occasion, budget reality, what it must never feel like, who it's secretly competing with. AI is excellent at asking these questions when told to; only you (and your client) can answer them.
In real work
"They asked things nobody else asked" wins pitches. The question list is the first thing a client sees of your process.
2 · Positioning: the sentence everything hangs on
For [narrow audience], [name] is the [category] that [genuine difference]. AI drafts these fluently — and almost all are soup. The de-souper: strike every word a competitor could also claim. What survives is your difference. If nothing survives, the brand doesn't have one yet — which is a strategy discovery, and billable.
3 · Systems, not assets
- Voice: 4 adjectives, each with a "this not that" pair ("playful, not wacky").
- Type: 2 pairings with rationale tied to the positioning — reject any rationale that could justify anything. Licence checked (Google Fonts = clear).
- Palette: named colours + where each is allowed. Usage rules are the system; swatches are decoration.
- Layout logic: grid habits, spacing personality.
4 · Copy variants without losing the voice
AI is a variant machine — controlled by a tone spec: your adjectives + this-not-thats + banned words. Without it: LinkedIn voice.
Watch out
Model copy loves "elevate," "seamless," "vibrant," "discover." Ban them by name. A good banned-words list is a brand asset.
5 · The stress test: critique loops for systems
Phone-screen size, sticker size, the support-slot poster where your name is 10% of the layout, the cynical Dublin gig-goer — and accessibility contrast. Then revise the single weakest element: a stress test that ends in a redesign was a tantrum, not a test.
The contrast trap
Models estimate contrast and confidently get it wrong. Check the two text pairs in the real calculator (WebAIM) and record the numbers.
6 · Production: where the hours come back
Asset checklists with specs, export QA lists, naming conventions, alt text drafts, style-guide assembly. Pure friction, fully delegable — with specs verified, because platform dimensions drift and models misremember them.
03This Week's Words
PositioningThe one sentence everything argues for: audience, category, difference.
Brand voicePersonality rules for all words: adjectives + this-not-that pairs.
Design systemThe rules keeping a thousand future assets coherent. The opinions are yours.
Type pairingTwo typefaces with assigned jobs and rationale tied to positioning.
Design tokensNamed values carrying the system into code and templates.
Usage rulesWhere each colour/type/element is allowed — the system's substance.
LockupA fixed arrangement of logo + text elements.
Contrast ratioThe accessibility maths of text on colour. Real tool, not vibes.
Campaign systemOne idea flexed across formats without falling apart.
Tone specThe voice rules you hand the model before asking for copy.
04Demo: Toy Cathedral — Weak Idea to Campaign Kit
I made this, so can you
"A logo and poster for my synth band" becomes a coherent system in four moves — with the human decisions marked.
Prepared demo artifact
Instructor backup / finished example: campaign-kit.html. Use it if the live tool run fails, or to reveal the finished structure after critique.
- Interrogation first. The model asks the 6 questions a brand designer would ask; the instructor answers with personality. The answers are the strategy.
- System, not assets. Positioning ×5 (one de-souped live, words struck with the room), voice + this-not-thats, type rationale (one rejected: "could justify any font"), palette with usage rules — [GUESS] marks visible.
- Stress test. Phone, sticker, support slot, cynic — weakest element revised, one targeted fix.
- The contrast moment. The model says the palette pair passes. WebAIM, live. The model loses the argument — and the habit installs itself.
05Step-by-Step Workflow: Micro-Brand System in One Session
- Interrogate: the model asks its 6 questions; you answer in writing. These answers are strategy.
- Position: 5 draft lines → de-soup → keep the one no competitor could claim.
- Voice: 4 adjectives + this-not-thats + 5 banned words.
- Type: 2 Google-Fonts pairings, each with rationale citing the positioning.
- Palette: 5 colours + usage rules + both text pairs checked in WebAIM (record the ratios).
- Stress test: run the attack; revise the single weakest element.
- Checklist: assets per deliverable with specs + export QA list.
- Assemble the sheet — and design it. The brand sheet for a brand is a brand artifact.
06Prompt Recipes
The interrogationI want "a logo and poster for [project]". Before you make anything: ask me the 6 questions a senior brand designer would ask first. Number them. Don't answer them yourself.
The system builderFrom my answers, write: (1) a one-line positioning — "for [audience], [name] is the [category] that [difference]"; (2) 4 brand voice adjectives, each with a "this not that" pair; (3) two type-pairing directions with rationale tied to the positioning — Google Fonts only; (4) a 5-colour palette with hex codes AND where each colour is allowed; (5) three layout concepts described in words. Mark anything that's a guess about my audience with [GUESS].
The de-souperHere are 5 positioning lines: [paste]. For each, strike every word that could equally describe a competitor. Show me what's left. Which line has the most surviving words?
The tone spec + copy variantsVoice: [4 adjectives + this-not-that pairs]. Banned words: elevate, seamless, vibrant, discover, [yours]. Write 8 headline variants for [deliverable + audience], in this voice, max 7 words each. Then mark which 2 you think are weakest and why — I decide the winners.
The stress testStress-test this brand system: [paste sheet]. How does it fail at phone-screen size? At sticker size? On a support-slot poster where the name is 10% of the layout? What would a cynical [audience member] say? Which palette pairs likely fail accessibility contrast for text? Identify the single weakest element and propose one targeted revision — not a redesign.
The production checklistBuild the asset/export checklist for [one concrete occasion]: every deliverable with dimensions/format/colour space, plus a pre-export QA list (spelling, contrast, bleed, file naming convention: [yours]). Flag any spec you're not certain is current.
07Lab: Brand System Stress Test (70 min)
Cold start friendly. No micro-brand? Standing brief: a late-night toastie hatch beside a Dublin venue, called "Notions."
- Interrogate + position (20 min): 6 questions answered in writing → 5 lines → de-soup → keep the survivor. ✏️ your answers.
- System draft (20 min): voice + banned words; 2 type pairings with rationale; palette with usage rules — contrast-check both text pairs in WebAIM and record the numbers.
- Stress test (15 min): run the attack; revise the single weakest element; note what and why.
- Checklist + sheet (15 min): production checklist for one occasion; assemble and design the mini brand sheet.
Zero-AI path (fully equal)
The interrogation and stress-test question lists are printed in this page — run them on paper. Google Fonts and WebAIM involve no AI at all. This week's lab never requires generation.
08Integrity Check
This week's four
- Count your ✏️ marks — are most decisions on the sheet yours? Under half: whose brand is this?
- Are [GUESS] labels still visible — or did guesses quietly become facts?
- Did you verify contrast with the real checker, not the model's estimate?
- Client work from this method: "AI used for drafts and variants; strategy, selection, and final design mine."
09Troubleshooting
| Problem | Fix |
|---|---|
| Positioning lines are all soup | De-soup harder. If nothing survives, the difference isn't real yet — that's strategy information, and it's valuable |
| Every type rationale sounds plausible | That's the tell. Reject any rationale that could justify anything; demand it cite the positioning |
| Copy sounds like LinkedIn | No tone spec. Feed voice + banned words; ban "elevate" by name |
| Model says contrast is fine | Models estimate; WebAIM knows. Check and record the ratios |
| The stress test praises everything | Anti-sycophancy clause (Week 2) + name the attacker: "a cynical gig-goer who's seen 1,000 posters" |
| Your system looks like every AI brand (gradient, geometric sans, teal) | You accepted defaults. Your taste enters through rejections — kill them and state what replaces them |
10Resources
- VideoLive Design Critique — The FuturEmbedded above. Critique culture from the professional side; long-form — dip in.
- ToolLinkedWebAIM Contrast CheckerThe real contrast maths from today's lab. Models estimate; this knows. Bookmark forever.
- ToolLinkedGoogle FontsFree, licence-clear typefaces — the lab's type source. Use the pairing suggestions, then make your own call.
- ArticleLinkedBrand strategy for designers — The FuturThe strategy-before-assets discipline this week applies, from people who teach it commercially.
- SiteLinkedBrand New — UnderConsiderationDaily professional identity reviews — taste training by volume. Partially paywalled; read the free ones critically.
- SiteLinkedIt's Nice ThatCreative-industry coverage including the ongoing AI-in-design debates — staying current without hype.
- ReferenceVerify before teachingPlatform spec references (Instagram / print dimensions)Specs drift constantly — verify current numbers the week you teach; never trust a model's recall of them.
Optional extension
- Flex test: apply your system to a deliverable you didn't plan for (tote bag, ticket stub, loading screen). Did the rules hold?
- Run the whole workflow on the standing brief in 25 minutes — speed reveals which steps you actually understand.
- Hard mode: write the rejection memo — 150 words explaining why you rejected the AI's most plausible direction. Defending a rejection is the purest judgment exercise in design.
Next week: UX and Product Design → — research synthesis, flows, wireframes, and the personas-are-hypotheses rule.