Meet us at AsiaBerlin Summit 2026, where our co-founder David Falk will be speaking. →

Learn more

4 — Brand Judgment

Article

How Does a Brand Make Decisions When Knowledge and Rules Are Not Enough?

Machine-readable is not enough. On the decision-making logic that turns brand knowledge into action when rules conflict or context changes.

Author
Tim Herzog
Tim HerzogCEO
Reading time
7 Minutes
Date
Abstract blue and pink gradient for Brand Judgment

When we talk about making brands accessible to AI, the conversation increasingly starts in the same place: brand knowledge needs to become machine-readable.

What is currently scattered across guidelines, campaigns, websites, assets, presentations and people's heads needs to be structured, connected and made understandable to machines. This is where the idea of a Brand Ontology comes in. Rather than simply turning a brand into another digital brand book, it describes the elements of a brand and the relationships between them so that a system can understand what belongs to the brand, where that knowledge comes from and in which context it matters.

That still leaves the question of what a machine should actually do in a particular situation. A system can have access to everything a brand knows about itself and still face several possible decisions.

A brand cannot be fully translated into rules

The obvious starting point is to make existing brand knowledge more precise. We define which colours can be used, how the logo should appear, which terms belong to the brand, which visual styles are appropriate and what tone of voice we expect. Tacit knowledge becomes explicit rules, and scattered documents become a connected system.

This works well as long as the task is unambiguous. It becomes more difficult when brand rules themselves require interpretation. A statement such as "We communicate confidently, but never arrogantly" is not a precise instruction for a human. It describes a space within which people can make different decisions depending on the situation.

For a machine, that statement raises further questions. What does confidence mean in a performance ad, and what does it mean in a CEO statement? Can a claim become more assertive if we know it performs better? How does that boundary shift in a different cultural context? And what happens when clarity, performance and tone of voice pull in different directions at the same time?

We could try to describe each of these situations with additional rules. But every new rule creates new combinations and edge cases. The attempt to codify a brand completely leads to an increasingly complex system without making every future situation predictable.

Brands are not made up of rules alone. They also emerge from the relationships between those rules and the way they are interpreted in different contexts. A brand wants to be recognisable without constantly repeating itself. It wants to remain globally consistent while being locally relevant. It needs to protect its identity while continuing to evolve. It wants to attract attention without becoming arbitrary, and it needs to deliver performance without trading long-term brand value for short-term results.

These tensions cannot be eliminated by writing a more comprehensive brand guideline. They remain because contexts, markets and brands continue to change.

Contradictions are not necessarily errors

The sources from which a machine learns about a brand are rarely free of contradictions either. Alongside the official brand guidelines, there are websites, campaigns, social content, product communication, local markets and countless decisions made over the years.

A guideline may define a particular tone of voice while the brand's communication has already evolved in another direction over the past two years. A successful campaign may deliberately break with an established visual principle. A local team may have developed an interpretation that works well in its market but has never become part of the global system. Meanwhile, management may already be working on a new direction for the brand that is not yet visible in any existing asset.

A knowledge system can try to resolve these differences by identifying the most recent or authoritative source and deriving a single truth from it. In many cases, that is necessary. In others, valuable information is lost in the process.

A guideline may describe what a brand aspires to be, while existing communication shows what it actually is. A new campaign may offer an indication of where the brand is heading. These statements do not necessarily cancel each other out.

A Brand Ontology therefore needs to understand the context surrounding a statement as well as the statement itself. Where did it come from? When was it created? Which part of the brand does it apply to? How authoritative is the source? How confident can we be in its interpretation?

This allows a system to distinguish between an error, an exception, an outdated rule and an actual change in the brand. A contradiction does not always have to be removed. It can remain part of the brand's knowledge.

Knowledge alone does not produce a decision

A Brand Ontology can know that two rules exist, understand their origins and contexts, and recognise that they are in tension with one another. None of this automatically determines which of several possible decisions should be made in a particular situation.

Until now, people have largely taken care of this step. Good brand leads and creative directors do more than know the rules of a brand. They weigh up which ones matter in a given situation, when an existing precedent provides guidance and when a new context calls for a new decision.

They can also deliberately depart from an established pattern without making the brand arbitrary. Such a departure may eventually become something that naturally belongs to the brand.

As Brand Agents begin to work more independently, part of this decision-making logic needs to be represented alongside brand knowledge. The aim is not to recreate human intuition. A system needs to understand how different principles relate to one another, which priorities apply in a given context and when the available information is insufficient to make a decision independently.

This introduces another task between knowledge and execution: choosing among several potentially valid options based on the context in which a decision is being made.

From Brand Rules to Brand Judgment

A rule-based system can know that a particular visual language belongs to a brand. A system with additional decision-making logic can also consider how binding that visual language is, the contexts in which it developed and whether a departure from it is possible within the brand's framework.

When addressing a new audience, entering a different cultural context or deliberately developing a new expression for a campaign, knowing that something matches the existing visual language is not enough. The system needs to relate that information to other goals and rules.

Beyond facts and rules, a machine-readable brand therefore needs information about how binding a rule is, when it applies, which other principles limit it, which precedents exist and which decisions people have made in the past.

This layer can be described as Brand Judgment. Not as another module alongside Brand Ontology, Brand Control and Brand Engineering, but as the decision-making logic connecting them. The Ontology provides brand knowledge and its relationships. Brand Control defines which actions fall within the permitted boundaries in a given context. Brand Engineering ensures that decisions and their outcomes flow back into the system.

Brand Judgment describes how existing knowledge becomes an action in a particular situation.

Autonomy becomes a question of context

This also allows us to think about autonomy in a more differentiated way. Rather than assigning an agent a fixed level of autonomy, each decision can be evaluated in its own context.

An agent may autonomously produce tens of thousands of variations of an established asset when it operates within a proven system and mistakes can be reversed at any time. The same agent should perhaps not autonomously publish a new global claim, because a single decision could have a far greater impact on the brand.

Relevant factors include novelty, confidence, reversibility and potential impact. Regulatory requirements and existing precedents may also influence whether a system acts independently, makes a recommendation or involves a person.

Autonomy and control are therefore not necessarily in conflict. The more accurately a system can assess the context of a decision, the more precisely we can determine where autonomous execution makes sense and where human judgment remains necessary.

Decisions become brand knowledge again

A decision can also change what the system knows about the brand. When a team deliberately departs from an existing rule in a new situation and that decision proves successful, it may establish a precedent. If similar decisions are repeated, a new principle may emerge and eventually become part of the established brand system.

This creates a cycle between Brand Ontology, Brand Control and Brand Engineering. The Ontology provides existing brand knowledge and its relationships. Brand Control defines the boundaries within which decisions can be made and executed. Brand Engineering incorporates new decisions, their outcomes and human feedback, allowing the system to develop over time.

Between these layers sits the decision itself: applying existing knowledge to a new context, weighing competing requirements and recognising whether an established rule is sufficient or a new decision is needed.

A machine-readable brand therefore does not need to describe every possible future situation in advance. It needs enough structure for machines to deal with situations for which no specific rule exists, while also recognising when their own basis for making a decision is insufficient.

Machine readability determines whether a machine can understand a brand. Brand Judgment determines how it makes decisions based on that understanding.

Share it