In late 2022 and through 2023, several major agencies ran controlled experiments: give a junior designer two hours with traditional tools and the same brief to an AI-assisted workflow. The AI workflow consistently produced more initial concepts — but the best final work still required a human hand to resolve the ideas.
Coca-Cola's "Create Real Magic" campaign (March 2023) became the most-cited real case. The company opened OpenAI's DALL-E and ChatGPT APIs directly to fans and agency creatives, inviting them to remix 100+ years of Coca-Cola brand assets. The campaign generated over 120,000 pieces in its first weeks — and surfaced dozens of directions the internal team had not considered.
Logo design follows a predictable funnel: brief → research → concepting → refinement → delivery. The concepting stage — generating dozens of rough directions before committing to one — has historically consumed 30–50% of total project time. It is also the stage where AI tools provide the clearest speed advantage.
Image generators like Midjourney, Adobe Firefly, and DALL-E 3 can produce 20–40 distinct visual directions in the time it takes a designer to sketch five. More importantly, they can explore stylistic territories — brutalist letterforms, Bauhaus geometry, organic hand-lettered marks — that a designer might not naturally reach for on a given brief.
The critical insight is that AI does not replace the concepting phase; it accelerates exploration. The designer's judgment — knowing which directions are legally defensible, culturally appropriate, and technically reproducible — remains the bottleneck.
WPP and OpenAI jointly built a custom platform giving agency creatives access to GPT-4 and DALL-E fine-tuned on Coca-Cola's brand archive. Designers could type a brief and receive compositions incorporating the Spencerian script logo, the red disc, and contour bottle silhouette. The campaign won a Cannes Lions Grand Prix in the Brand Experience category — the first AI-assisted campaign to do so.
When Wolff Olins senior designers discussed the Coca-Cola campaign publicly in 2023, they noted a consistent pattern: AI concepts needed human curation, not just selection. The tool would produce a direction with the right mood but the wrong geometry, or the right palette but a form that would not scale to a 16×16px favicon. A designer's trained eye was still the gate.
The practical workflow that emerged across agencies in 2023–2024 looks like this:
Adobe Firefly (integrated into Illustrator's Generative Recolor), Midjourney v6, DALL-E 3 via ChatGPT, and Stable Diffusion with ControlNet are the four most-cited tools in agency concepting workflows as of mid-2024. Each has distinct strengths: Firefly respects commercial licensing; Midjourney produces the most aesthetically polished outputs; ControlNet allows structural control over generated marks.
Understanding the limits is as important as understanding the capabilities. As of 2024, AI image generators consistently struggle with:
Exact typographic control — generating a specific wordmark with correct letterform construction is unreliable. Text in AI images is notoriously distorted.
Trademark clearance — an AI has no mechanism to check whether a generated mark is confusingly similar to an existing registered trademark. This remains an entirely human legal responsibility.
Vector output — all major image generators produce raster files. A logo must ultimately exist as a scalable vector. Vectorization is a separate, manual step.
Brand strategy — AI cannot determine whether a logo direction is strategically correct for the client's competitive positioning. That requires business analysis and client knowledge that lives outside the model.
A client is launching a specialty coffee subscription brand called Meridian. They describe their brand as: "precise, globally curious, warm but not folksy." They want a logomark (icon only) and are open to abstract or representational forms.
In this lab, work with the AI to develop and refine prompt language you would use with Midjourney or DALL-E to generate initial concept directions. Ask about prompt structure, style descriptors, and how to guide the AI toward specific moods or forms. Complete at least 3 exchanges.
In 2023, several Pentagram partners publicly discussed using AI tools — including Adobe Firefly and custom Stable Diffusion workflows — to rapidly prototype brand system components: pattern libraries, color palette stress-testing across hundreds of simulated applications, and type combination previews. The work that received the most attention was their rebrand of The Museum of Arts and Design (MAD) in New York, where AI-assisted pattern generation was used to explore visual motifs drawn from the museum's collection.
The key finding: AI was most useful not in creating the final system elements but in pressure-testing them — generating hundreds of simulated real-world applications (billboards, tote bags, digital banners) to reveal where the system broke down before any assets were produced.
A logo alone is not a brand identity. A complete brand identity system typically includes: the primary logomark, wordmark, and lockup variations; a defined color palette with primary, secondary, and neutral tones; a typography system (display, body, UI typefaces); a pattern or texture library; an iconography style; photography and illustration direction; and a tone of voice guide.
AI tools have become useful across nearly all of these components — at different levels of maturity and reliability.
Color is one of the highest-leverage AI applications in brand identity. Huemint (launched 2022) was among the first tools to demonstrate that a machine learning model trained on successful brand palettes could generate contextually appropriate color combinations from a text brief.
The documented professional workflow as of 2024:
Jones Knowles Ritchie's 2021 Burger King rebrand — a return to the 1969–1994 visual identity — was completed before AI tools were widely available, but it has since become a benchmark case study for testing AI-assisted brand system generation. In 2023, design educators at RISD and Parsons used the BK brief as a controlled test: asking students to rebuild an equivalent system using AI tools. The AI-assisted cohort produced comparable breadth in half the time but required more senior designer review to match strategic precision.
Type pairing is a domain where AI provides genuine utility. Tools like Fontjoy use neural networks to suggest complementary typefaces based on similarity and contrast scores. Adobe Fonts has integrated AI pairing recommendations into its platform since 2022.
The limitation is specificity: AI tools work from large populations of font data and tend to recommend mainstream pairings. For brands that need a distinctive or unexpected typographic voice — the kind of pairing that becomes a brand signature — human type directors still produce more differentiated results.
A hybrid approach has emerged: use AI to generate 20–30 pairing candidates quickly, then apply human editorial judgment to identify the one pairing with genuine character.
The brand guidelines document itself — the PDF or Notion page that codifies the identity system — is increasingly being structured with AI writing assistance. ChatGPT and Claude are used to draft the "tone of voice" sections, write usage rules in clear language, and generate the "do/don't" examples.
Importantly, Figma's AI features (released 2023–2024) allow designers to auto-generate component variants and apply brand tokens across an entire design system, dramatically reducing the mechanical production work in building a comprehensive guidelines document.
AI excels at generating the raw material of a brand system — color options, pattern motifs, type candidates, mockup visualizations. It does not determine which materials are strategically correct for the brand. That judgment — rooted in competitive analysis, audience insight, and cultural context — remains the designer's core value.
You're developing a brand identity for Hawthorn — a direct-to-consumer herbal wellness brand. Brand attributes: natural, clinical precision, quiet authority, approachable expertise. Primary audience: health-conscious adults 28–45.
Use this lab to work through color palette selection and typography system decisions with AI guidance. Ask the advisor to help you evaluate specific color combinations, discuss type pairing logic, or test your decisions against the brand attributes. Complete at least 3 exchanges.
In August 2022, computer scientist Stephen Thaler attempted to register a copyright for an image generated entirely by his AI system "DABUS." The US Copyright Office rejected the application, ruling that copyright requires human authorship. Thaler appealed. In February 2023, the Copyright Office issued formal guidance confirming the position: works produced entirely by AI — without human creative selection or arrangement — are not copyrightable.
The same month, graphic artist Kris Kashtanova received a partial copyright for the graphic novel "Zarya of the Dawn" — but only for the human-authored text and arrangement. The individual AI-generated images were stripped from the copyright registration. This established a working precedent with direct implications for brand designers using AI tools.
As of 2024, US copyright law (and largely parallel frameworks in the EU and UK) treats AI-assisted brand design work under three categories:
The Copyright Office's March 2023 guidance explicitly stated that it would evaluate AI-assisted works "on a case-by-case basis" to determine whether the human contributions are "sufficient to constitute authorship." For brand designers, this means documenting the creative decisions made during an AI-assisted process — which prompts were written, which outputs were rejected, what modifications were made — creates a defensible record of human authorship.
The trademark question is distinct from copyright and, for most brand designers, more commercially urgent. Trademark protects a mark's function as a brand identifier — its ability to distinguish one company's goods from another's.
The US Patent and Trademark Office (USPTO) does not (as of 2024) require that a trademark applicant own copyright in the mark. An AI-generated logo that is distinctive, is used in commerce, and does not conflict with existing registered marks can be registered as a trademark.
The risk for brands using AI-generated logos: distinctiveness. If an AI model generates a similar mark for two different clients — which is statistically possible given the finite output space of current models — both brands could attempt to register the same or similar marks, creating costly legal conflicts.
A separate and ongoing legal question concerns whether AI-generated images that closely resemble existing copyrighted works constitute infringement. The Getty Images v. Stability AI lawsuit (filed January 2023 in the UK and US) is the most prominent active litigation on this question. Getty alleges that Stable Diffusion was trained on Getty's copyrighted image library without license.
For brand designers, the practical implication: when an AI generates a logo concept that closely resembles a known brand mark, that output carries legal risk regardless of the AI's training data. A prompt that yields something visually similar to the Nike Swoosh is a problem — not because of how it was generated, but because of how it would function in the marketplace.
This reinforces the professional standard: all AI-generated brand concepts must go through trademark clearance search before any client presentation, just as a hand-drawn logo would.
Leading IP attorneys and brand consultancies recommend a four-step AI brand asset protocol: (1) Document all human creative decisions in the AI workflow for copyright defensibility. (2) Run all AI-generated concepts through a professional trademark clearance search before client delivery. (3) Ensure any commercially delivered logo is either substantially human-modified (Category 3) or registered as a trademark rather than relied on for copyright protection. (4) Review this process against your jurisdiction's evolving case law — this area is changing rapidly.
Adobe explicitly addressed the training data concern when launching Firefly in 2023. Adobe stated that Firefly was trained exclusively on Adobe Stock images, openly licensed content, and public domain works — creating what Adobe calls "commercially safe" AI outputs. This is why Adobe Firefly has become the preferred AI tool in many agency workflows: clients accept it more readily because the licensing provenance is documented.
It does not resolve copyright questions about the output itself, but it significantly reduces the risk that a Firefly-generated concept is a direct derivative of a specific copyrighted work.
A startup client has asked you to design a logo for their new fintech app. They want to use Midjourney to generate the icon, and they want to claim full copyright in the resulting work. They've also asked whether they need to do a trademark search.
Use this lab to work through the legal questions with an AI advisor. Ask about copyright eligibility, what documentation to maintain, trademark search requirements, and how to advise the client. Complete at least 3 exchanges.
Superside — a subscription-based creative services company with over 700 designers globally — publicly repositioned as an "AI-first" creative agency in 2023. Their documented approach: integrate AI tools at every stage of the brand identity process, from initial brief analysis through concept generation, with human designers handling curation, refinement, and client relationship management. They reported a 40% reduction in concept-phase time across brand identity projects.
The client communication challenge they identified was significant: some clients initially perceived AI involvement as lower quality or less personalized. Superside's response was to lead with outcomes — showing the breadth of concepts delivered — rather than process. The number of clients who objected to AI use dropped substantially when they saw the delivered work first.
Based on documented workflows from agencies including Superside, Huge, Collins, and independent studios, the following tool stack represents current professional practice in AI-assisted brand identity work:
Red pills = AI generation tools. Gold pills = production / refinement tools. The professional workflow passes through both categories in sequence — AI for generation and exploration, traditional design tools for refinement and production.
Industry surveys from AIGA and Design Week (2023–2024) document the following shift in brand identity project timelines at studios using AI tools:
The pricing question is the most contested in the industry. Three models have emerged in documented agency practice:
Value-based pricing: Price is set by the strategic value of the brand identity to the client's business — not by the hours invested. AI efficiency gains go to the designer's margin, not discounted to the client. This is the model advocated by most senior brand consultants.
Deliverable-based pricing: Price is set per defined deliverable (primary logo, secondary mark, color system, type system, usage guidelines). AI makes these deliverables faster to produce; the price per deliverable may shift downward over time but remains tied to outcomes, not time.
Hourly/day-rate with AI transparency: Some studios disclose AI tool costs as a separate line item, similar to stock photography or font licensing. This model tends to be used by studios with clients who have procurement requirements around AI use disclosure.
A widely circulated thread on Brand New (the brand identity criticism blog) in mid-2023 documented a dispute between a client and agency over whether AI-assisted work should cost less. The client's argument: "if it took less time, it should cost less." The agency's counter-argument: "we're charging for the judgment, strategy, and experience that guided the AI — not the hours." The thread received over 200,000 views and no consensus was reached. The debate remains active in the industry.
Based on documented agency experience, three client types have emerged with distinct communication needs around AI use in brand design:
AI-enthusiastic clients want to know exactly which tools were used and how. They often have unrealistic expectations about speed and cost. Communication priority: set realistic expectations about the human work that still dominates project time.
AI-skeptical clients worry the work will feel generic or lack personality. Communication priority: demonstrate the human curation and strategic judgment through the process — show the rejected directions, explain why choices were made.
AI-indifferent clients (the majority) care about outcomes, not process. Communication priority: lead with the work. Disclose AI use in the contract and process documentation, but don't make it the conversation's center of gravity.
AI changes the speed and economics of brand identity design. It does not change what brand identity design is for: building a coherent, distinctive, legally defensible visual identity that connects a business to its audience. Every AI tool decision should be evaluated against that purpose — not against the tool's novelty or capability in isolation.
You've completed a brand identity project using an AI-assisted workflow (Midjourney for concepting, Adobe Firefly for color exploration, Illustrator for vector production). The work is strong. Now you're presenting to two different clients in back-to-back meetings: one is AI-enthusiastic and asking detailed questions about tools; the other is AI-skeptical and worried the logo "looks like a robot made it."
Use this lab to practice how you'd handle both conversations. Ask the AI advisor how to frame your process, respond to specific objections, or discuss pricing rationale. Complete at least 3 exchanges.