Most companies are asking AI to create more content.
The Coca-Cola Company is exploring a more consequential question:
How can AI control the way thousands of creative assets are produced—without weakening one of the world’s most recognizable brand identities?
Its answer is Project Fizzion, an AI-powered design system co-developed with Adobe. Instead of using generative AI simply to produce more images, Fizzion turns Coca-Cola’s brand rules into machine-readable instructions that can guide creative work as it happens.
This represents a significant shift in enterprise AI.
AI is moving from being a content generator to becoming part of the infrastructure that governs how content is created, adapted and approved.
From a Red Swatch to Machine-Readable Brand Rules
Coca-Cola’s struggle for brand consistency predates artificial intelligence.
According to Fast Company, former Coca-Cola president Robert Woodruff reportedly carried a sample of the company’s distinctive red in his wallet during the 1930s. When he encountered Coca-Cola branding, he could compare the colour with the official sample.
Today, that problem exists on an entirely different scale.
Coca-Cola sells approximately 2.2 billion servings per day through around 30 million points of sale. Its marketing ecosystem spans more than 200 brands, over 200 countries and numerous languages, channels, agencies and product formats.
A physical colour sample—or even a conventional brand manual—is no longer enough.
Coca-Cola’s visual guidelines reportedly extend to hundreds of pages. Every regional team and agency partner must correctly interpret rules covering typography, logo placement, spacing, imagery, colour and composition.
The problem is not the absence of rules.
It is ensuring that those rules are interpreted consistently thousands of times across different markets and formats. Fast Company
What Project Fizzion Actually Does
Fizzion is designed to convert static guidelines into an active design system.
It operates as an intelligent cloud service connected to plugins inside Adobe Creative Cloud applications, including Illustrator, InDesign and Photoshop. Designers continue working in familiar tools while the system learns from the decisions they make.
As designers build layouts, adjust typography and position visual elements, Fizzion captures their creative intent and encodes it into what Coca-Cola calls a StyleID.
A StyleID is a machine-readable representation of how a brand or campaign should look and behave. Once trained, it can guide the production of new layouts, sizes and localized variations while applying the relevant visual rules.
In practical terms, the workflow looks like this:
- A designer creates an approved campaign direction.
- Fizzion learns the relationships between its logo, typography, imagery and layout.
- These decisions are encoded in a StyleID.
- Teams and agency partners use that StyleID to produce campaign variations.
- Brand rules are applied while the content is being created—not only during a final manual review.
Adobe says the system can help teams generate hundreds of variations, including resized layouts and adaptations for different platforms and regions. Adobe’s Project Fizzion interview
This makes a StyleID less like a traditional brand document and more like executable design logic.
A PDF explains what people should do.
An intelligent design system helps the tools apply those instructions directly.
Why This Is More Than Content Generation
Generative AI has made content production dramatically easier. A marketer can produce images, copy and variations within minutes.
But that creates a new enterprise problem: the faster content is generated, the harder it becomes to govern.
Every additional asset creates another opportunity for:
- Incorrect logo placement
- Inconsistent typography or colours
- Outdated campaign elements
- Poorly adapted formats
- Unnecessary review cycles
- Brand dilution across markets
Fizzion addresses this problem upstream.
Instead of waiting for a brand manager to identify mistakes after an asset has been created, the system is intended to apply approved design behaviour during production.
That changes AI’s role.
It is no longer only helping people create. It is helping an organization maintain standards while creation scales.
Coca-Cola’s Broader AI Evolution
Fizzion also shows how Coca-Cola’s approach to generative AI has matured.
In 2023, the company launched Create Real Magic, an experimental platform developed with OpenAI and Bain & Company. It allowed digital artists to create original work using GPT-4, DALL·E and Coca-Cola’s archived brand assets.
That project focused largely on experimentation and creative participation. The Coca-Cola Company
Fizzion addresses a different stage of adoption.
The progression is revealing:
- Experimentation: What can generative AI create?
- Integration: How can AI fit into professional design tools?
- Governance: How can AI-generated and AI-assisted work remain consistent?
- Scale: How can approved creative systems be deployed across markets?
This is the path many large organizations are likely to follow. Once the novelty of generating content fades, the harder questions concern quality, ownership, governance and operational control.
What “Up to 10× Faster” Really Means
Coca-Cola says Fizzion can enable creative teams to produce content up to ten times faster. At its May 2025 announcement, however, the company described Fizzion as being in its pilot phase. The Coca-Cola Company’s Fizzion announcement
The speed claim should therefore be interpreted carefully.
Neither Coca-Cola nor Adobe has publicly provided a detailed benchmark explaining:
- Which production tasks were measured
- What the previous workflow required
- How many campaigns or users were evaluated
- Whether the improvement was typical or a best-case result
- How much human review remained necessary
The claim is plausible for repetitive work such as resizing layouts, adapting approved campaign visuals and creating format variations. These tasks can involve substantial manual production and repeated approval cycles.
But “up to 10× faster” does not mean every campaign will be completed ten times faster. It is currently a company-reported performance claim rather than an independently verified result.
Designers Are Still Central—But Their Role Changes
Coca-Cola and Adobe describe Fizzion as a designer-led system: designers establish the creative direction, while AI helps repeat and adapt their decisions.
That distinction matters.
AI is effective at reproducing patterns and enforcing defined constraints. It is far less reliable at deciding when a brand should break from convention, respond to cultural change or develop an entirely new visual language.
Fizzion therefore does not eliminate the designer. It potentially shifts the designer’s work from manually producing every variation to defining the system from which variations can be created.
The designer becomes:
- A creator of the original concept
- A teacher of the system
- A curator of acceptable outputs
- A manager of exceptions
- A guardian of creative evolution
The repetitive work may decrease, but responsibility for defining what “on-brand” means becomes more important.
Brand Consistency Is Not the Same as Complete Compliance
Fizzion’s public documentation focuses mainly on visual identity: logos, typography, imagery, layouts and campaign adaptations.
There is currently no public evidence that it automatically validates every legal requirement, advertising claim, product fact, accessibility rule, copyright issue or country-specific regulation.
A visually correct advertisement can still contain an inaccurate claim, culturally inappropriate message or legally inadequate disclosure.
Human review therefore remains essential—particularly for regulated content and culturally sensitive campaigns.
The strongest model is not uncontrolled automation. It is machine-enforced visual standards combined with accountable human judgment.
The Risks Coca-Cola Will Still Need to Manage
Turning brand guidelines into software creates powerful efficiencies, but it also introduces new risks.
Incorrect rules can scale quickly.
If a StyleID contains an error or outdated instruction, that problem could be reproduced across hundreds of assets.
Consistency can become uniformity.
A system optimized to repeat established patterns may discourage unusual ideas and make campaigns feel overly standardized.
Local relevance cannot be reduced to layout rules.
Colours, images, gestures and messages can have different meanings across cultures. Visual compliance does not guarantee cultural relevance.
Dependence on one technology ecosystem can grow.
Because Fizzion is deeply integrated into Adobe’s products and services, Coca-Cola may become increasingly dependent on Adobe’s infrastructure, pricing and technical direction.
Performance claims require transparent measurement.
Speed alone is not enough. Coca-Cola must also measure quality, effectiveness, error rates and the amount of human correction still required.
How Fizzion’s Real Value Should Be Measured
The most meaningful evidence will not be the number of assets generated. It will be whether the system improves the entire content supply chain.
Useful measures would include:
- Time from approved concept to market-ready assets
- First-pass approval rate
- Number of brand violations detected
- Rework and correction hours
- Cost per localized asset
- Time spent on formatting versus creative development
- Performance of localized content
- Number of exceptions requiring human intervention
If Fizzion merely produces more content, it is another automation tool.
If it reduces production delays and errors while preserving creative quality and local relevance, it becomes a genuine operating system for global brand management.
What Other Companies Can Learn
Coca-Cola’s approach suggests that companies should not begin by asking, “Which AI tool can generate our content?”
They should first ask:
- Which decisions are repeated across our content workflow?
- Which brand rules can be expressed clearly?
- Where do errors and approval delays occur?
- Which tasks require human judgment?
- How will we measure quality as output increases?
Companies do not necessarily need Coca-Cola’s scale to apply this principle.
A smaller organization can begin by creating structured brand components, approved templates, reusable prompts, defined review criteria and clear human-approval checkpoints.
The objective is not maximum automation.
It is controlled scalability.
The Bigger Shift
Coca-Cola’s most important AI innovation may not be its ability to generate another advertisement.
It may be its attempt to transform decades of human design knowledge into a system that software can understand and apply.
That is the deeper significance of Project Fizzion.
The first phase of generative AI made content abundant. The next phase will determine whether organizations can manage that abundance without losing quality, trust or identity.
The competitive advantage will not belong only to companies that generate content fastest.
It will belong to those that can scale creativity while keeping every execution recognizable, relevant and accountable.
The future of AI in marketing is not simply creation. It is governed creation.
If helpful, I can set up “Monitor Project Fizzion updates” so you can revise the pilot status and performance claims when Coca-Cola publishes new evidence.
Leave a Reply