Brand work, as it is practised by most agencies in 2026, is a category in slow decline. The work has become so codified — discovery interviews, mood boards, three-route presentations, a guidelines PDF — that the customer's experience of buying it is roughly identical regardless of which agency they choose. The deliverable is increasingly mistaken for the brand itself, and the actual job of brand work — building a system that can survive being operated by the people who will operate it — is, in our experience, almost universally undertaken as an afterthought.
We have spent most of the last decade running brand engagements alongside engineering engagements, and the contrast has changed how we think about the work. A brand system is software in a different medium. It has a runtime — the people who apply it. It has a contract — the situations it is meant to handle. It has a maintenance burden — every new application stretches or violates the original system in ways the original guidelines did not anticipate. Treating brand work as a one-shot deliverable, rather than as the foundation of a small operating system, is the structural reason most brands erode within eighteen months of launch.
The deliverable is not the brand
The most common artifact of a brand engagement is a guidelines PDF — sometimes a sprawling Notion site, occasionally a Figma library, but typically a 60-to-200 page document that is intended to govern every future application. The document is, in almost every case, too long to be read by the people who will apply it and too short to be exhaustive against the situations they will face. It serves a procurement function — it justifies the cost of the engagement — but it does not, on its own, govern the brand in the way it is intended to.
What governs the brand, in practice, is the small number of decisions the operators have already memorised — the wordmark, the type pairings, the colour palette, the photography direction, the voice. Everything beyond that is, sooner or later, made up by the operator on the day. The honest job of a brand engagement is to make sure those small decisions are good ones, that they are durable under the conditions they will be applied in, and that they can be recalled without consulting a document.
Brand systems are operated, not consumed
A brand system is operated by a small number of named people inside the customer's organisation — often a single in-house designer, sometimes a marketing manager with no design training, occasionally an external agency on a per-project basis. The system has to be designed for those people. A system that requires a senior brand strategist to apply correctly is a system that will be applied incorrectly within a month of handover.
The discipline this imposes is a strong constraint on complexity. We refuse to ship brand systems that contain primitives the operating team cannot produce themselves — custom illustration styles that depend on a specific external illustrator, photography directions that require a budget the customer does not have, motion languages that depend on a tool the team does not own. Every primitive in the system has to be operable by the team that has been named as its owner. If a primitive is not operable, it is not part of the system, regardless of how good it looks in the launch deck.
A brand is not what the agency ships. A brand is what survives the second year of being operated by the team that inherited it.
What a small operating system looks like
The brand systems we ship now are deliberately small. A wordmark, a small type pairing — usually one display serif and one neutral sans — a controlled colour palette of no more than six values, a photography direction expressed in three reference images and a single sentence, a voice document of no more than 300 words, and a motion language reduced to three named families. That is the entire system. Everything beyond it is left to the judgement of the operator, with the system primitives strong enough that even a poor judgement call cannot do too much damage.
This is, in our experience, the configuration that survives. The customer can recall the entire system from memory within two weeks of handover. New marketing materials can be produced without consulting a guidelines document. The brand is recognisable across surfaces because the primitives are simple enough that they cannot be misapplied without conscious effort. The system grows over time — the operating team adds primitives as they need them — but it grows from a foundation that the team owns, rather than from a foundation that an external agency periodically returns to refresh.
Why most brand engagements over-deliver in the wrong way
Most brand agencies are paid in proportion to the volume of the deliverable. The procurement process rewards the agency that ships the most artifacts — more guidelines pages, more brand applications in the launch deck, more bespoke components in the design library. The customer feels they have received more for their money, and the agency feels they have justified their fee. Both feelings are sincere. Both produce systems that are too large to be operated.
The fix is to change what the engagement is paid for. We charge for the durability of the system, not for the volume of the artifacts. The launch deck is small. The guidelines document is short. The library is intentionally constrained. The cost of the engagement is justified by the system's behaviour over the eighteen months after handover, which is a harder thing to demonstrate in a procurement conversation but is the only metric that actually matters to the customer over the lifetime of the brand.
Brand work and engineering work, on the same page
We run brand engagements alongside engineering engagements because the disciplines reinforce each other. A brand system that has been engineered for operation by a small team is the same kind of artifact as a software system that has been engineered for operation by a small team. The constraints are the same — limit the primitives, document the failure modes, design for the people who will run the system rather than for the people who specified it. A brand that survives is, in this sense, an engineering deliverable. The earlier the brand and the engineering teams agree on this framing, the better the system that ships.