Machine-readable version of this page (for AI agents): /what-is-brand-engineering.md
Contact

Brand Engineering Is an Old Idea

What brand engineering means, where it comes from, and how we do it.

Vertumnus by Giuseppe Arcimboldo, a portrait made of fruit, vegetables, and flowers, with the face cut out in cream stock and set in numerals
Supermega Design, after Vertumnus by Giuseppe Arcimboldo, 1590–91.

TL;DR Brand engineering turns a brand into a system that people and agents can build with. The name is new. The idea goes back at least to 1964. It matters more now because agents make a lot of the work, and an agent only knows the brand you give it.

A New Name

This year we kept hearing the title "brand engineer." In the US, Google searches for "brand engineering" went from 50 in August to 140 in September and stayed there in October. Those are small numbers, but they nearly tripled in a month.

Tommy Geoco of UX Tools visited the design teams at Vercel, Perplexity, Metalab, Ramp, and Snowflake. On the Dive Club podcast he told a story from Vercel. A brand designer there had never written code. The marketing team published a lot of blog posts, every post needed a card image, and the design team made the cards. So the designer built a small tool. Anyone in marketing dropped in an image, and the card came out with, in Geoco’s words, "all of the branding and the sizing baked in." After that, the team started calling him a brand engineer.

We build tools like it for our clients’ brands. For Readog, an iOS app that helps people read more, you type the title of an article and get three blog headers made from the brand’s own parts. We wrote about how that brand was made in Brand Design With Opus 5.5. Other tools we’ve built generate social media posts, turn illustrations into vector files, and resize one visual for every format a brand ships in.

The Readog blog header tool. Type an article's title, and it builds the header from the brand's parts and checks it against the rules.

One Name, Three Meanings

The people who use the term don’t agree on what it means. We found three meanings.

The first is marketing operations. The site brandengineering.ai describes one person with AI agents doing the work of a marketing team: the research, the articles, the social posts, the newsletter, and the reporting. Chris Riley, who wrote the site’s essay on the role, says his agent produced 537 social posts, 10 articles, and eight newsletters in 47 minutes. The team at Defyner calls the system around a company’s agents a "brand harness," and the discipline that builds it brand engineering. In this meaning the brand is one input, and the work is building the pipeline.

The second is brand knowledge for AI. The company brand.ai says, "Brand engineers turn your brand into machine-readable rules." Column Five, a content marketing agency, defines brand engineering as "treating your brand identity as a living system that is present at the moment of creation," and lists four things that system has to carry: messaging, visuals, taste, and intelligence. Most of those rules are about how the brand writes.

The third comes from designers, and the Vercel story is the plainest example of it. A designer takes the brand’s layouts, type, and sizes, and builds them into a tool other people can use. Two studios describe the same work without the name. Instrument wrote in March that a brand now lives in the systems that express it, and it publishes small tools to test that. Wild keeps each client’s brand as plain text files in the client’s own repository.

All three agree that a document can’t hold a brand once many people and many tools are making things with it, that the brand has to exist in a form a tool can load, and that people still make the calls. As brand.ai puts it, "A brand engineer can’t build a system that operates by itself."

In the first two meanings somebody else has already designed the brand, and the engineering starts from their guidelines. We are a design studio, so we mean the third. We design the brand, and then we turn it into a system other people can build with.

The Idea Is 60 Years Old

In 1964 the Swiss designer Karl Gerstner published a book called Designing Programmes. Its subtitle, in English translation, is "Instead of solutions for problems, programmes for solutions." He wanted designers to make the system that produces many answers, and then choose among them. "The creative process is to be reduced to an act of selection," he wrote. The Museum of Modern Art gave the idea an exhibition in 1973.

Two years after the book, New York’s Transit Authority hired the design firm Unimark to advise it on the signs in the subway. Bob Noorda and Massimo Vignelli designed a modular system, with separate components for arrows, route markers, and train information, and three sizes of type. According to the design historian Paul Shaw, the Transit Authority didn’t have the money for a full manual, so its own sign shop built the signs from the drawings. The shop ignored the modules and lettered the type by hand. The drawings also showed a black stripe at the top of each sign, which stood for the metal channel that holds a sign up. The shop took it for part of the design and put it on the signs. Vignelli called the result "the biggest mess in the world."

The manual came in 1970. By then the stripe was on the signs, so the manual made it part of the standard.

We recently rebuilt that manual as a Figma library, page by page. The design was good, but for four years the sign shop had no version of it they could build from.

Signs from the 1970 New York subway manual, rebuilt in Figma. On the left, as the manual drew them. On the right, as they were built.

Then we built a tool on top of the library. Poster Factory generates posters in the manual’s system. You generate a set, pick a style, and tune one poster. If the poster breaks one of the system’s rules, the tool won’t let you download it.

Poster Factory: 32 posters in six styles, built from the 1970 subway sign system.

Try Poster Factory

After the manual, some identities came with tools. In 2007 Sagmeister’s identity for Casa da Música, the concert hall in Porto, came with a color tool that samples any image and pulls 17 colors from it for the logo. In 2011 the new identity of the MIT Media Lab was an algorithm with 40,000 possible logos, enough to give each person at the lab their own. Pentagram built a shape generator for the chip company Graphcore so that, as Pentagram describes it, Graphcore’s "internal team" can "create infinite patterns."

Around 2014, Jon Levine and Jina Anne at Salesforce named the design token. A token stores one design decision, a color or a spacing value, under a name, and every product reads it from there. A shared file format for tokens reached its first stable version on October 28, 2025.

The manual, the generators, and the token file each gave other people a version of the design they could build from. Brand engineering is the current name for that work.

What Changed

These days, a lot of what a brand puts out is made by AI agents.

Agents are fast, and without direction they all make the same thing. We wrote in March that AI "always goes to the mean." Anthropic says the same about its own models. Ask one for a landing page without guidance, and "it will almost always conform to Inter fonts, purple gradients on white backgrounds, and minimal animations." Anthropic calls the cause "distributional convergence." A model picks the choices that appear most in its training data.

A 2024 study in Science Advances measured this. In the study, 293 people wrote short stories, and some of them got story ideas from GPT-4. The stories written with AI ideas were rated higher, and they were also more similar to each other. A 2026 review of 19 studies, covering ideas, designs, writing, and visual art, found that work made with AI comes out more alike. The effect was small and statistically significant.

Figma said it plainly when it opened its canvas to agents: "AI agents haven’t had this context, which is why so many designs created by AI often feel unfamiliar and generic." By context it means the decisions a team has already made, from color palettes and button padding to typography.

So the design tools now give agents the team’s own design system. Through Figma’s MCP server, agents from outside Figma build with a team’s components and variables, and in November Figma said it was its fastest-growing product last quarter, by weekly users. Claude Design "checks its output against your design system." Google open-sourced a file format called DESIGN.md so that agents "can know exactly what a color is for."

An agent builds whatever you ask for, right away, from whatever you give it. If you give it only a logo and four colors, it fills in the rest with the average.

Supermega Design, after Napoleon Crossing the Alps by Jacques-Louis David, 1801.

How We Do It

We build each brand two systems: a visual system and a brand content system.

The Visual System

The visual system holds every design element from the brand work, each built as a component. Each element is drawn once as a vector file and comes in the variants the brand uses, so an agent can place it and resize it without redrawing it. The elements are sorted by kind and named, so a person or an agent can find the right one. Their values, such as colors, type sizes, and spacing, are stored as design tokens, so every file and tool uses the same ones. Every kind of asset the brand makes gets a template built from those elements, with its sizes and layout already set.

Supermega Design, parts of the MyVault brand.

Designing a brand means deciding on color, composition, type, and patterns. We write each decision down as a rule, including what the brand never does. A rule can say which colors go together, how much empty space a visual keeps, or how many elements it can hold. The agent reads the rules before it makes anything and checks its work against them afterward. A person approves the result.

The Brand Content System

The brand content system is the mind of the brand. It teaches an agent to write and sound like the brand.

Both systems grow as the brand needs new things.

The Tools

The main tool is the agent, working in Figma with the parts and the rules. Someone on the client’s team asks for an asset, and the agent builds it and checks it against the rules. Around the agent we build visual tools for the assets a team makes most, such as a generator for social media posts and one for blog headers. You type a title or drop in an image, and you get a finished, on-brand visual. Other tools turn illustrations into vector files or check finished work with a vision model.

The Guides

We write two. People get a page in Figma that shows the parts, the templates, and finished examples. Agents get a skill, a set of instructions they load before a task.

For Readog, this system produced more than 200 finished visuals in three days, and the client’s team can keep making them after we leave.

A person still chooses the direction, edits the output, and says no, and most of our time now goes there. Gerstner wrote of his own method that "the work is not diminished; it is merely transferred to another plane."

Small Teams and Large Teams

A small team still needs blog headers, social posts, and ads every week, and it rarely has anyone free to make them. With the system, the marketer asks the agent for a header or a week of social posts and gets finished visuals built from the brand’s own parts. If the team has a designer, they look over the work and fix what’s off, and they don’t have to produce every piece themselves. The team puts out a lot more than it did before, and all of it looks like the same brand.

In a bigger company, designers, marketers, regional teams, and outside agencies all make things with the brand, and now their agents do too. Each of them reads the guidelines a little differently, and after a year the brand looks different in every channel. When everyone builds from the same parts and the same rules, the drift stops, and the checks catch off-brand work before it goes out. A new hire or a new agency opens the guide and can make something usable right away. The design team gets fewer small requests and fewer review rounds, and has more time for new work.

What Brand Engineering Is

Brand engineering turns a brand into a system that people and agents can build with.

A brand project used to end with guidelines, a PDF or a Figma file. Everyone who made something new had to read it and rebuild the brand by hand. Now it ends with a system. The team and its agents use it to make new work.

The people who design a brand should also engineer it. They know it best. Every rule in the system comes from a decision made during the design, and the designer knows why each one was made. Someone who only gets the guidelines sees the rules but not the reasons, and doesn’t know what to do when a new asset doesn’t fit them.

The system is never done. When the brand needs a new kind of asset, someone adds the parts and rules for it. From then on, the agent can make that asset whenever someone asks. The client’s team can do this themselves, with the same tools we used.

If you want this for your brand, book a call and tell us what it has to make next.

Questions

What Is Brand Engineering?

Brand engineering turns a brand into a system that people and agents can build with.

What Does a Brand Engineer Do?

A brand engineer builds the brand’s parts as components, writes its rules down, and makes the small tools and templates that produce on-brand assets. The title is new and mostly informal. At Supermega Design the people who design a brand also engineer it.

Is Brand Engineering New?

The name is recent, and people use it in at least three ways: for marketing operations run by agents, for brand rules written for AI, and for a brand built into parts and tools. The idea is older. Karl Gerstner argued for designing "programmes for solutions" in 1964, and studios have shipped identity tools since at least 2007.

How Is Brand Engineering Different From Brand Guidelines?

Brand guidelines are a document. They show the logo, colors, type, and examples, and every time someone makes something new, they have to read the document and rebuild the brand by hand. Brand engineering turns the same decisions into working parts: components in Figma, rules an agent can follow, and tools that produce finished assets. You start from the brand’s own parts instead of a blank page.

How Is Brand Engineering Different From a Design System?

A design system is part of it. A design system is a library of components and rules, usually for a product’s interface, and people still put each screen together themselves. Brand engineering starts from a library like that and adds rules an agent can follow, tools that make finished assets, and checks that catch off-brand work. It also covers the whole brand, not only the product.

By Markiian Bobyliak

Markiian is the founder and design director at Supermega Design.

More from the blog

Brand Design With Opus 5.5 Brand Content System in Claude Code