The first living, AI-driven identity
Statistics
- Client
- ADC
- Deliverables
- Data-driven branding
- Generative visualizations
The challenge
As artificial intelligence continues to reshape industries, ADC is strengthening its positioning as a trusted leader in an increasingly complex and abstract field. This evolution demanded a brand identity that could move beyond static systems and familiar tech clichés.
The challenge was to create an identity capable of evolving alongside both AI and ADC itself: technically credible and analytically rigorous, yet warm, expressive, and human. It needed to remain coherent across diverse contexts while accurately reflecting ADC’s culture, values, and expanding body of work.
Value delivered
We created the first living, AI-driven brand identity for ADC, built around transformation and collective intelligence. Starting from analytical and mathematical symbols, the system continuously generates organic, expressive patterns through a custom-trained AI model that understands ADC from the inside out, its people, expertise, and ways of working. The model learns how and when different expressions are appropriate, refining the identity across hundreds of real-world applications, from reports to storytelling moments.
The result is a brand system that redefines authorship and longevity. Shaped not only by designers but by ongoing input from the ADC team, the identity evolves over time while maintaining a shared visual language. It balances precision with emotion, demonstrating how AI can amplify creativity rather than standardize it. The outcome is something visually unlike what the industry has defaulted to: not sterile, not generic, but expressive and distinctly human in character. A future-ready identity that grows with ADC and its people.
Background
We began by auditing ADC's previous identity, which felt static and disconnected from their technical depth, and identifying the need for a system that could better reflect their evolving culture. Through workshops and research, we unpacked their language, values, and foundations, mapping how ADC thinks, communicates, and positions itself within the industry. We explored their history, the expertise of their people, and the ways their work sits at the intersection of data, technology, and human insight. From that foundation, through rapid prototyping and continuous team feedback, we developed a generative system designed to learn and adapt over time.
Phases and toolkit items used
Visualstyle
While the core of the identity is driven by algorithms and code, its expression is designed to feel warm, inviting, and unmistakably human.
We built a visual system that feels optimistic, confident, and alive, a deliberate contrast to the sterile, technical aesthetics that have come to define the AI sector, and a reflection of the people and culture at the heart of ADC.
Logo
The ADC logo isn't a separate entity. It emerges from the shape language and custom typography that runs through the entire identity system. Rather than sitting apart from the brand, it feels like a natural extension of it: the brand’s static stamp in a system that is otherwise always in motion. It holds its ground without competing with the patterns around it. To ensure contrast against the vibrant, expressive patterns, the logo is used exclusively in black or white.
Color
We deliberately moved ADC away from the dark, tech-oriented aesthetic that dominates the sector. The palette is bold and vibrant, featuring violet, pink, teal, green and gold.
These aren't static swatches; the colors are designed to move and shift across the generative patterns, creating a sense of energy and variation that changes with context. Together they reflect a brand that is confident, curious, and unafraid to challenge convention.
Typography
To complement the generative shapes, we selected Ease by Studio Feixen. Neither rounded nor straight, the typeface embodies a state of transition, sitting somewhere between structure and expression, much like the identity itself. It's a typographic nod to the latent space in AI models: that in-between territory where meaning takes shape. Quirky yet refined, technically credible while remaining warm and approachable, it carries the same balance the rest of the identity strives for.
Patterns
The identity consists of patterns that are technically precise but imperfect and expressive. Mathematical symbols, the native language of ADC's domain, morph based on proximity, density, and context, creating compositions that feel organic even though they're entirely computational. There’s an intentional looseness to it: the system is precise in its logic but allows for variation, imperfection, and surprise in its outputs. Rooted in data and driven by algorithms, but never cold.
Transformation
The identity begins with the raw ingredients of ADC’s work: data, AI, and mathematics. We turned fundamental operators and formulas into the building blocks of the visual system — simple symbols that hold deep complexity. Through algorithmic repetition and proximity logic, they morph and merge into evolving geometries, mirroring how intelligence forms through connection and transformation.
At the core sits a generative tool powered by more than 50 parameters, translating meaning into composition. Quadtrees structure the spatial logic beneath the surface, creating rhythm and coherence across every asset. The result is infinite variation on a shared foundation: every expression is unique, yet all are driven by the same computational logic.
Aliving,evolvingsystem
To create an identity that doesn’t just look like ADC but thinks like ADC, we built a system that generates visuals from meaning rather than randomness. At its core is a custom AI model trained on more than 100 data points about ADC’s work, values, and people — forming a living knowledge base that informs every output, so each expression carries context as well as form.
Underthehood
The system is built on embeddings: numerical representations that map relationships and meaning across a high-dimensional space, where content naturally gravitates toward related concepts. A finance report settles near other reporting documents; a culture update lands closer to people and team context.
The critical step was connecting those 100 ADC data points to visual output. Through a manual mapping process, we assigned specific patterns to each one, teaching the model to translate context into composition.
For every input, the model produces three variations derived from slightly different neighbors within the embedding space. When a user selects their preferred output, that choice feeds back into the system, refining the model over time. This is what makes the identity truly alive, continuously shaped by the people using it.
Patternevolution
As the people at ADC interact with the system, generate new patterns, and make choices, that input feeds back into the model and shapes what comes next. The more it is used, the more it learns. The same financial report cover, generated at different points throughout the year, will yield different results based on what the model has learned.
The result is the world’s first truly living brand identity, one that evolves through the hands of the people behind it.
Technicaldetails
Trainingdata
Contextual data about ADC
We built a custom dataset representing ADC by collecting over 70 public sources and documents. These were distilled into 100 contextual data points that capture the company's work, culture and people
Userinput
- 384 dimensional vector
- all-MiniLM-L6-v2 embedding model
The user input is turned into a vector, a point in 384 dimensional space, using the sentence-transformers all-MiniLM-L6-v2 embedding model.
Similaritysearch
- PGVector (PostgreSQL)
In the training data we search for neighboring vectors which are close to our input vector using cosine similarity. We can configure how many similar vectors and patterns we are able to retrieve.
Patternfusion
Temperature-scaled interpolation
These neighbors are then fused into three variations. Each uses a different blending temperature, controlling how much the nearest versus furthest neighbors influence the result.
Feedbackloop
Reinforcement Learning from Human Feedback (RLHF)
The user selects their favorite, which is marked as eligible training data and becomes available as a neighbor for future queries. The system evolves through implicit user preference learning.
Livingidentity
Preference-driven dataset evolution
Each generation pulls from previously chosen patterns across all users. Over time, this shared selection history feeds back into the neighbor pool, shaping output. Collective preference gradually steers the visual language.
Thetool
The generative tool we built is web-based and accessible to everyone at ADC. Users start by providing context: a project description, a blog post, the brief for an upcoming campaign. The tool then places this content in ADC's embedding space, mapping it against the company's knowledge base. Based on where it lands, the tool generates three variations, each a slightly different interpretation of the same input.
The user picks their favorite, which is fed back into the model to refine future outputs. Under the hood, the tool runs on a custom algorithm built in JavaScript, adhering to the design rules and parameters established during the mapping process. The result can be exported as vector or image across a range of formats, ready to use in presentations, documents, or social media. No design skills needed, just context in, identity out.
Livingbrandelements
The evolving nature of the identity unlocks possibilities that traditional brand systems cannot offer. Every interaction generates new brand data, preferences accumulate, and the system becomes more attuned to the people and context behind it. This dynamic is most visible in the Collective Map: a live visualization of every pattern generated across ADC, forming a continuous real-time feed of the identity in motion. From technical reports to cultural moments, each output leaves its mark on a shared canvas.
Brandvitalitydashboard
The brand dashboard gives ADC a live window into how their identity behaves over time. It visualizes usage statistics, shifting preferences, and collective choices across the team: which patterns are being generated, which are being selected, and how individual parameters are trending.
More than a reporting tool, it reflects the living nature of the system itself, showing how a brand can evolve through the everyday decisions of the people using it. A real-time overview of an identity that never stands still.
Identityinuse
Results
By building a living, generative identity for ADC, we created something that has never existed before: the world's first identity system that truly evolves. Driven by a custom tool that translates the company's knowledge, values, and people into a unique visual language, it stands apart from the cold, generic aesthetics that dominate the AI sector. Every visual choice, from the mathematical symbols it starts from to the expressive patterns it produces, reinforces that contrast. Complex technology made accessible, human, and engaging. And with every use, the identity grows, ensuring the brand transforms alongside ADC as they transform the way organizations adopt AI.
