Brand Management Becomes Infrastructure: Why Brands Must Become Machine-Readable in the AI Era.

Brand infrastructure refers to a brand’s structured, versioned, and verified knowledge; the interfaces through which people and AI systems access it; and the guidelines that define what may be said and done on behalf of the brand.

In April 2025, “Sam,” the AI support agent for the code editor Cursor, responded to a user who was constantly being logged out when switching between devices. According to Sam, Cursor was intended for use on only one device per subscription for security reasons. No such rule existed; the bot had made it up (CX Network). According to reports, users canceled their subscriptions. It wasn’t until about three hours later that the company clarified on Reddit that no such rule existed; it later apologized and has since labeled AI-generated support responses (eWeek).

This case is instructive in two respects. First, the agent spoke with the authority of the brand: his human-sounding name and the tone of a corporate information service made the response sound like an authoritative statement. Second, it wasn’t a lack of intelligence, but rather a lack of a verified source for this statement and an authority to review it before it was released. It’s impossible to tell from the outside what information Sam had at his disposal. Furthermore, this was a matter of fact rather than opinion—in other words, a straightforward case.

It has become a given that agents are aware of themes, tone, and no-gos. The question is, where do they get this information, and who is responsible for it? Brand guidelines are a genre with a long history, written for designers, agencies, and employees. They assume a reader who interprets the text, brings context to bear, and asks questions when in doubt. This assumption no longer holds true once systems generate text, images, and interfaces and increasingly prepare or make decisions on their own.

My thesis is this: Brand management must therefore shift its focus. If brands are also applied by machines in the future, brand knowledge will no longer serve merely as a guide for creatives, but as infrastructure. Brand implementation would then no longer be just a question of how consistently a brand is implemented, but of how it is operated. A concept such as the “Spherical Brand”—which views a brand through nine spheres: Sustainable, Purposeful, Responsible, Humane, Conversational, Archetypical, Evolving, Intelligent, and Living—can serve as a useful framework for understanding this. An introduction to this concept is available at thinkmoto.de.

What the Guidelines Do Not Include

Four gaps explain why traditional guidelines are insufficient for automated application.

1. Appearance over decision. Guidelines describe how a brand looks and sounds. But agents must decide: who they target, what they reject, and when to hand it off to a human.

2. Implicit rather than explicit. Michael Polanyi described this as “tacit knowledge ”: We know more than we can put into words. A large part of brand knowledge lies in experience, approval processes, and the minds of long-time employees. For an agent, only what is written down and accessible actually exists.

3. Many copies instead of a single source. Robert Haase described here on tbobm how Copilot, Brand Kit, Agenturdeck, and Brand Book each contain their own version of the same brand, with each version becoming outdated at a different rate (the brand as an MCP server). I agree with this assessment. In my view, the copies are the symptom; the cause is the absence of a single source with an owner.

4. Not verifiable. Neither a person nor a machine can reliably apply a rule that lacks a verification criterion. “Authentic” is a value, but not an instruction.


Brand as Infrastructure?

Sociologist Susan Leigh Star has described infrastructure as something that is embedded, remains invisible during normal operation, and only becomes visible when a disruption occurs. Brand knowledge behaves exactly the same way today. As long as people implicitly correct it, its gaps go unnoticed. As soon as systems access this brand knowledge—for example, to produce content in bulk without oversight or even to make more far-reaching decisions—those gaps become incidents.

The term is therefore more than just a metaphor. It defines what a brand must deliver in order to be trusted:

Characteristic of InfrastructureWhat This Means for Brand Awareness
A single, unambiguous sourcePeople and systems draw from the same, well-maintained repository rather than from copies. However, this repository must be equally unambiguous for both to interpret. (This requires a translation process.)
Defined InterfacesBrand information is available for use in applications, not just as a PDF file
VersioningIt is clear which version of the trademark was in effect and when
VerifiabilityRules include evaluation criteria and justifications; decisions are documented
Operation and ResponsibilitiesSomeone manages the portfolio and is responsible for what happens on behalf of the brand


Not all of these characteristics are a matter of technology. The last one is a matter of organization. Infrastructure is not a software purchase, but rather a mode of operation.

What Needs to Be Made Legible

The more difficult question is: What exactly should be made machine-readable? Color values, claims, and spellings can be easily encoded. Attitudes cannot, because they manifest themselves in judgments, and those must be documented differently than rules.

A legal stance can rarely be expressed as a specific legal provision, but it can be illustrated through case law. This can be compared to legal practice, where legal norms only take shape through adjudicated cases. This is an analogy, not proof, but it explains why a purely rule-based approach falls short in this context.

A brand infrastructure therefore requires, in addition to a set of rules, a growing body of documented decisions that illustrate how the brand weighs its options. For rules, a checklist, rationale, and escalation process are sufficient. For brand stance, positive and negative examples and clear-cut borderline cases are needed.

Consistency, Coherence, and the Address of the Judgment

In his article, Robert Haase distinguishes between consistency—that is, verifiable facts—and coherence—the judgment of whether something fits the brand. Servers that provide brands to agents deliver rules but do not make decisions. His statement on this: “The judgment has no address.”

I consider this to be the key point and would frame it in architectural terms. A server that delivers rules is part of the infrastructure, but not its core. Three functions are needed, all drawing on the same knowledge base:

Intelligence consolidates brand knowledge, identifies gaps, and monitors changes in the business environment.

Creation generates text, images, and dialogs that are based on this knowledge rather than on a generic prompt.

The Governance team reviews content prior to publication, manages approvals, and documents who made which decisions and when.

Governance is where decisions are formalized. An archive stores what has already been approved. An infrastructure verifies content before it is approved. The shared core—which we call Brand Core—contains the brand’s operational logic, which all three functions access. In Spherical Brand AI, we implement this structure: Brand Memory and Due Diligence form the Intelligence; Writer, Image Studio, and Assistants form the Creation; and Design Check, approvals, and a documented approval trail form the Governance. This principle applies regardless of the tool used.

Robert's counterpart, “What the Machine Doesn't Get, ” adds to this: What a brand deliberately withholds from machines is also brand knowledge. An infrastructure must therefore document exclusions and limits just as much as it does approvals.

What This Means for Brand Implementation

Those who view brand knowledge as infrastructure transform brand management in four key areas.

1. From enforcement to maintenance. Brand management is becoming, in part, a form of knowledge curation: brand knowledge has an owner, an editorial process, and a release cycle.

2. From guidelines to case files. Every escalation, every instance of “That doesn’t fit with who we are,” is knowledge. It belongs in the case file—with an explanation—rather than disappearing into an inbox.

3. From spot checks to protocols. Governance tracks what has been reviewed, corrected, and escalated. From this, the brand learns where its rules are unclear.

4. Authority becomes explicit. Whoever gives an agent tools grants them authority. Haase calls this “Agent Authority.” It requires a name, a date, and boundaries.

This also shifts the self-image of those who implement brands. In “A Day Without an Agency, ” Erich Posselt argues that agencies will have to sell judgment and context rather than production in the future. A brand infrastructure is a way to make this judgment permanently available, even when the person who made it isn’t in the room at the moment.

A practical starting point doesn’t require a platform. Gather the ten decisions from the past three months in which someone stopped or changed a publication, and write down the reason for each one. This list serves as the raw material for rules, examples, and review questions, and it shows just how much brand knowledge has existed only in people’s minds until now.

Borders

Not everything can be formalized. Polanyi’s insight applies here as well: a certain element remains implicit, and an over-formalized brand becomes rigid. Spherical Brands are conceived as learning systems; their knowledge must therefore remain adaptable and modifiable. When it comes to major decisions, humans remain responsible, and questions remain open regarding standards and liability when an agent acts on behalf of the brand. An infrastructure is no substitute for judgment; it provides a place and a memory for it.

Conclusion

Brand guidelines have traditionally been written for people who can ask questions. Since machines don’t ask questions, the knowledge must be available in advance: structured, verified, and with a reference point for judgment. This isn’t a matter of tool selection, but rather one of operational approach. Brands that establish this early on remain recognizable as themselves even when machines are working on their behalf.

Sources

Haase, Robert: " The Brand as an MCP Server: How a Brand Responds When an Agent Asks a Question," tbobm.com

Haase, Robert: What the Machine Can't Get, tbobm.com

Posselt, Erich: A Day Without an Agency, tbobm.com

Polanyi, Michael (1966): The Tacit Dimension. Doubleday, Garden City

Star, Susan Leigh (1999): “The Ethnography of Infrastructure.” *American Behavioral Scientist* 43(3), 377–391

CX Network (Amelia Brand, April 24, 2025): Chatbot hallucination at Anysphere sparks customer backlash and subscription cancellations

eWeek: Cursor AI Chatbot's "Hallucination" and Fake Policy

Spies, Marco: What Is a Spherical Brand? The Framework for Brand Management in the AI Era ((Brand New Thinking, May 15, 2026)

7. November 2026
A post by:
Marco Spies

Marco Spies berät seit über 25 Jahren Marken. Er ist Mitgründer und Strategy Partner von think moto, einem Berliner Studio an der Schnittstelle von Marke, Experience Design und künstlicher Intelligenz, und hat für zahlreiche führende Marken gearbeitet. Sein Schwerpunkt ist Markenführung im KI-Zeitalter. Gemeinsam mit Katja Wenger schrieb er das Buch Branded Interactions. Lebendige Markenerlebnisse für eine neue Zeit. (Hermann Schmidt Verlag, 2018). Sein Konzept The Spherical Brand beschreibt Markenführung für eine Zeit, in der ökologische, technologische und kulturelle Veränderungen zusammentreffen. Mit Spherical Brand AI entwickelt think moto eine Markeninfrastruktur, die Markenwissen strukturiert, Inhalt markenkonform generiert und Markenverantwortliche bei der täglichen Steuerung der Marke unterstützt. Auf seinem Substack nowhere but now schreibt er Essays über Kultur, Macht und Technologie.

Comments are closed.

crossarrow-rightarrow-up-circlearrow-left-circlearrow-right-circle