AI explores the look; a designer makes it real
Visual identity is the part of branding where AI is most seductive and most misunderstood. It can generate a hundred logo-ish images in a minute, and almost none of them are a usable logo. The right frame is this: AI is for direction and exploration - aligning a team on where the look is heading - while the final craft, the production art, and the ownership stay with a human designer. Tonight you build Cadence's direction, not its finished identity.
Direction, not final craft 6 min live
AI image tools - Midjourney, DALL-E, Canva - are exploration engines. They let you and a team see logo feel, colour palettes, type mood, and overall vibe fast, so you can react and align before a designer spends a day. What they do not do is produce final, ownable, production-ready brand art. That handoff, and the ownership that comes with it, stays human.
LiveWhat AI is genuinely great at here - and where it stops3 min▶
Keep the line bright between exploration and production:
- Great at: generating mood, colour, and type directions at volume; building a moodboard to align a team; showing a designer where you are heading before the brief.
- Not for: final logo files, real vector art, a working type system, or anything you would trademark and ship. AI output is a reference, not a deliverable.
- The reason: a logo needs to be exact, scalable, and legally yours. AI raster images are none of those - and their provenance is unclear, which matters for the rail in Part 3.
Self-studyWhy strategy has to come before the look2 min read▶
The Clay 2026 guide is blunt about sequence: strategy and personality come first, then identity is the designed expression of them. That order is what stops AI visual exploration from becoming a pretty-picture generator. If you have done b2 and b3, you can judge an AI image by one test - "does this look like what Cadence is?" - instead of "is this cool?". On-strategy beats pretty every time, and it is the only way to steer image tools off their own version of the average.
Building visual direction with AI 5 min live
Good AI visual exploration is not random prompting - it is a recipe. The strongest moodboard prompts name five things: subject, style, palette, mood, and references. And every one of those should be tied back to Cadence's personality, so what comes out is on-strategy, not just pretty.
LiveThe five-part moodboard recipe, tied to personality3 min▶
Lock a prompt pattern so every image is comparable and on-brand:
- Subject - what is in frame (an app moment, a calm workspace, an abstract mark).
- Style - illustration, photography, minimal geometric, and so on.
- Palette - drawn from personality: Cadence is calm and capable, so soft, low-noise colour, not neon.
- Mood - the feeling: present, unhurried, quietly confident.
- References - describe a genre or era, never "in the style of [living artist]" - that is a rail-three problem in Part 3.
Self-studyHow AI helps with each visual element2 min read▶
| Element | How AI helps (direction only) |
|---|---|
| Logo feel | Explores mark styles and shapes to react to - never the final vector you would ship. |
| Colour | Generates palette directions tied to the personality; a human locks exact, accessible values. |
| Type | Suggests a type mood (geometric, humanist, serif); a designer chooses real, licensed fonts. |
| Moodboard | Assembles a cohesive visual world fast to align the team and brief the designer. |
| Imagery style | Explores photo/illustration directions - with commercial-use terms checked before any reuse. |
Every row is direction. The word "final" never appears - because final craft and licensed assets are the designer's, and the exact values are locked by a human.
The IP + likeness minefield - read this twice 4 min live
This is the loudest rail in the course, and it is not optional. AI images sit on top of three separate legal exposures at once - copyright, likeness, and trademark - and any of them can turn a beautiful direction into a liability. Nothing you generate here is adopted until a human, and where the stakes are real a lawyer, has cleared it.
LiveThree exposures, live at the same time2 min▶
- Copyright. AI models train on existing work, and prompting "in the style of [a living artist]" can reproduce protected expression. Ownership of raw AI output is itself unsettled - you may not own what you generate.
- Likeness / right of publicity. Generating a recognizable real person, or something close to one, risks a right-of-publicity claim. Do not put real faces in brand imagery.
- Trademark. An AI image can echo an existing logo or protected mark closely enough to infringe. A mark has to be cleared, exactly like a name in b4.
Self-studyWhat "rights-clear inputs" actually means2 min read▶
Rights-clear means the prompt and any uploaded reference do not lean on someone else's protected work or identity. Practically: describe genres, eras, and moods rather than named living artists; do not upload a competitor's assets or a real person's photo as a reference; and read what the tool actually grants you for commercial use, because it varies by plan and changes often. When in doubt, treat the image as a reference for a designer, not a deliverable - a reference carries far less risk than a shipped asset. This session covers the risk; it is not legal advice, and real filings and disputes go to qualified counsel.
Build a Cadence moodboard from a locked recipe ★ 13 min · everyone builds
Turn the five-part recipe into a real moodboard for Cadence, all tied to its calm, capable personality - then curate down to the few images that actually say what Cadence is.
Write your prompt from the five parts - subject, style, palette, mood, references - with every choice justified by Cadence's personality from b3.
Generate a set in your image tool. Keep the recipe fixed and vary one element at a time, so you can tell what each change does.
Curate ruthlessly to 6-9 images. The test is not "is it pretty" but "does this look like what Cadence is" - calm, present, quietly confident.
Run the rail as you curate: no living-artist styles, no real faces, and a note of the tool's commercial-use terms. Anything that fails the check comes off the board.
Assemble the survivors into one moodboard. This is direction to hand a designer - not Cadence's finished look.
The moodboard that aligned a team in an afternoon. A brand spent weeks arguing about "the vibe" in words. One locked-recipe moodboard - eight images, all on-strategy - settled it in a single meeting, because the team could finally point at what they meant. The designer then made the real identity. AI did not design the brand; it ended the argument so the designer could.
Write the visual-direction brief for a designer ★ 8 min · build your own
The moodboard is only useful if it hands off cleanly. Write a brief that tells a designer exactly what AI explored, what to keep, and what to make final - with the IP and likeness check built in so nothing questionable slips into production.
Summarize what the AI exploration surfaced - the palette direction, type mood, and overall feel that fit Cadence.
State clearly what to keep (the direction) and what stays the designer's to make final (logo, real type, exact colour values, production files).
Add the IP and likeness check: confirm no living-artist styles, no real faces, no near-copies of existing marks, and that any imagery used commercially is rights-cleared.
Send the brief plus the moodboard to a designer. You have given direction and protected the brand - the craft and ownership are now in the right hands.
Try it yourself - this week ◐ 30-45 min total
- Finish Cadence's moodboard from the locked recipe if you did not live, and curate it to 6-9 on-strategy images.
- Write the one-page visual-direction brief, with the IP and likeness check as a real section - not an afterthought.
- Read the commercial-use terms of the image tool you actually use. Note in one line what it does and does not grant you.
- Run the "does this look like what the brand is" test on three AI images for your own brand. Keep the on-strategy ones; delete the merely pretty.
- Optional: list any place your team has treated AI imagery as final. Flag it to revisit in the b10 governance session.
Official sources covered
Taught from public brand-identity guidance and the IP canon. Certificates and paid tooling stay official; legal filing stays with counsel. This session covers:
Three questions before you go 🎯 ◐ 90 seconds
1 · What is AI's real role in visual identity?
AI explores mood, colour, and type direction fast so a team can align. Final, ownable, production-ready craft stays with a human designer.
2 · What makes a moodboard prompt on-strategy rather than just pretty?
The five-part recipe keeps images comparable and steered by personality. On-strategy beats pretty, and it is how you push image tools off their own average.
3 · Which legal risks do AI brand images raise?
AI images sit on three exposures together: training-data copyright, real-person likeness, and near-copies of existing marks. Check commercial terms, use rights-clear inputs, and keep a human and legal review before adoption.