Explored a semi-autonomous brand brain with Vimeo, presented by Superside at Figma Config

In this talk, Phillip Maggs explains how we treat design systems and brand data as machine-readable intelligence — adding a structured, semantically mapped layer that AI can actually understand and work with meaningfully.

In short, we pilot a semi-autonomous email machine that writes personalized on-brand copy and wraps it into adaptive email templates, generated using the brand's design system components.

What I think is particularly interesting about this project is that, innovation emerged from looking backwards and reflect on how we build brands; and start wondering what will be the necessary requirements for a brand, in a future where LLMs and agents will be at the core of any user interaction.

What we came up with was a fundamentally different way of representing brand and design data, in a manner that language or diffusion models can easily access and retrieve from.

Knowledge graphs are sort of an ancient concept in the tech world, but they regained significance in the advent of AI because of their unique ability to store unstructured data (like brand voice, personality and other conceptual elements) together with value-based data (such as layout rules, for example). The combination of the two, brings more nuance and character to generative outputs. I believe that this will remain true for while, at least.