From designer to system architect: Designing a node-based brand identity system that structures human-AI co-creation in brand identity development

Abstract

Type a prompt, wait a few seconds, and generative AI hands you a finished logo. A logo belongs to a brand’s identity, but so do its strategy, personality, and how it shows up across touchpoints. Brand identity is a network of connected decisions, where strategy shapes personality, personality shapes visual language, and visual language shapes every touchpoint a customer meets. Current AI tools work against this logic. The brand strategy defined in one tool is invisible to the image generator in another, since they are designed for single, disconnected, prompt-based outputs. At each stage, brand decisions must be manually re-entered. As a result, designer control is lost, brand consistency is threatened, and the creative process begins to represent what the tool can do rather than what the brand should be.

This thesis investigates how a node-based brand identity system can structure human-AI co-creation across the full brand identity workflow while maintaining the designer’s strategic and curatorial role. The work follows a Research through Design approach, structured by the Double Diamond. An exploratory phase, combining a literature review, a focus group with four agency brand designers, and a design hackathon, mapped how branding is practised and where AI currently fits. These findings, together with five concepts drawn from brand theory and human-AI interaction research, formed the theoretical framework for the design. Guided by this framework, a node-based system was built in Krea.ai across the six core branding phases, from research and discovery to guidelines and delivery, with around 90 nodes.

The system was evaluated through a qualitative usability study with six professional designers, combining a guided walkthrough, a think-aloud assignment, semi-structured interviews, and a Likert-scale survey. Reflexive thematic analysis produced four themes. Designers found the phase-based logic sound and instantly recognisable, located the friction in the platform interface rather than the node-based architecture, and consistently reclaimed authorship by curating outputs, editing prompts, and uploading their own work. Every participant judged the designer’s role to be equally or more important when working with the system, not less.

The findings show that a node-based system shifts the designer from asset creator to system architect: someone who designs the structure that produces brand assets, not only the assets themselves. This shift is evidenced at two levels. Building the system was itself a design activity in which the decisive choices concerned architecture, which nodes to connect, which models to assign, and where to place human control points, rather than individual visual outputs. In the evaluation, designers who had not built the system took up curatorial, directive, and prompt-authoring control, added nodes themselves, and reasoned about its structure, showing this orientation is accessible to those who operate the system, not only the one who built it. This makes design expertise more decisive, not less, while raising the value of new skills such as turning creative intent into precise prompts. Because the framework structures any multi-phase creative process where strategic intent must be carried across stages, the same logic may extend beyond branding to other design domains.