Note on examples: The billing system anecdote in the opening section is a working example used across several notes and essays in this series. If you have already encountered it, you can skip to the section heading "The Semiotic Ladder."

Meaning in Enterprises Is Rarely Established

Many enterprises build systems without agreement on what a "customer" is, what "current" means, or how to represent a reversed transaction. The consequences are real; they show up in reports, audits, and downstream systems that stakeholders do not fully trust because the meaning was never made explicit before implementation, or even after.

In a billing system I worked on, a stakeholder reused a fee rate for a different purpose without documenting the change or informing anyone who relied on it. As a result, two teams ended up with two different meanings for the same attribute. The accounting team spotted the issue through their monthly reconciliation process and I traced the root cause by examining the configuration tables. An AI operating at scale without human oversight would apply a single interpretation across the data and produce incorrect revenue figures without a corrective mechanism. There would have been no means to identify the issue without knowing that net revenue should equal gross revenue in this narrow context. The revenue would be incorrect, and the issue would not surface until the payment arrived at a coarser grain that made automated reconciliation impossible. The failed reconciliation would have eventually forced a manual review with last month's books being reopened, and the business operating on an inaccurate understanding of its revenue during that time.

The Semiotic Ladder

Ronald Stamper developed the semiotic ladder during his work on organizational semiotics, extending the traditional semiotic divisions of syntax, semantics, and pragmatics, drawn from Charles Morris, by adding physical world, empirics, and social world levels. The ladder runs from the physical world at the bottom through empirics, syntax, semantics, and pragmatics, up to the social world of commitments at the top. Jan Dietz incorporated Stamper's ladder as one of three theoretical foundations for DEMO, Design and Engineering Methodology for Organisations, alongside Habermas's Communicative Action Theory and Bunge's Ontology.

The six levels, from bottom to top:

Physical The physical medium: electrical impulses, sound waves, marks on paper. No informational structure, no meaning, just physical substance and causal interaction.
Empiric Encoding, i.e. how signals are expressed: Roman letters, Morse code, binary digits. These patterns represent signals, independent of what they mean.
Syntactic Structure and rules: grammar, protocols, data formats. We can distinguish well-formed expressions from malformed ones, but there is still no meaning, only formalism.
Semantic Meaning: signs become associated with concepts; symbols refer to things, whether real or modeled. This is where most discussions of semantic technology claim to operate.
Pragmatic Intent: the question shifts from what something means to why it is being said and what action or commitment it implies.
Social Shared conventions, norms, roles, and mutual commitments. Meaning is now embedded in a community of practice, not just an individual interpretation.

Dietz groups these six levels into three categories: formaThe implementation side of communication, i.e. physical carriers, encoding, and structure., empirics and syntax, at the bottom, informaThe meaning-and-intent side of communication, i.e. semantics and pragmatics., semantics and pragmatics, in the middle, and performaThe level of social commitment, where a statement or act changes what people are committed to., social commitment, at the top. Most systems that claim to operate at the semantic level are actually operating at the syntactic level, parsing formats, validating schemas, checking structural consistency, without crossing into meaning at all.

Stopping at the semantic level is not enough: meaning without intent is a proposition, not a thought, and a proposition without commitment produces no change in the social world. This is Dietz's answer to why information systems that are semantically well-specified can still fail to guide action; they stop at informa and never reach performa.

AI agents do not escape the ladder; they occupy specific positions on it. Current AI systems, including large language models, operate primarily at the informa level; they process meaning, generate propositions, and can simulate intentional responses, but they cannot hold performa-level authority, because performa requires accountability for social commitments, which only human actors can bear. Intent must therefore be established and held by a human; the agent acts within that intent, not as its source. The consequences of mislocating an agent on the ladder, treating informa-level processing as if it were performa-level commitment, are amplified at scale, because an agent acting without human performa oversight produces no accountability trail when its outputs are wrong. The implications for how AI agents should be positioned within an organisation are addressed in Intent-Driven AI Delegates.

What Semantics Is Not

In practice, validity shows up in use, correction, and context, not in the specification itself. Drift happens in the space between major initiatives, in the everyday business-as-usual changes where nobody stops to ask whether the meaning of "customer" or "partner" has shifted since the last time anyone looked.

This working note addresses semantic practice in enterprise information systems and AI governance. Formal semantics, such as propositional logic, first-order logic, description logics, and communication protocols, is out of scope here.

Peirce: Meaning Is Triadic

Charles Sanders Peirce held that meaning is not a two-way relationship between a word and a thing but triadic: it arises from the relationship among sign, object, and interpretantThe effect a sign produces, such as an action, a response, or a further sign that carries the meaning forward., where the interpretant is the effect or next consequence produced by the sign. Without this interpretant, there is no meaning.

The interpretant is what the sign produces as its next consequence: a feeling, an action, or a further sign, which can itself produce further consequences.

Three examples:

The Community of Inquiry

Peirce extended this further with fallibilismThe view that any belief or interpretation may turn out to be wrong and therefore remains open to correction. and the concept of the community of inquiry: meaning is stabilized over time through a community of inquirers who test, challenge, and correct interpretations against experience. Without a correction mechanism, signs drift from their objects.

In his philosophy of science, the community of inquiry is explicitly deliberate: scientists formulate hypotheses, design experiments, publish results, challenge each other's conclusions, and revise. The correction mechanism is institutionalized and intentional.

But Peirce also thought inquiry has a natural, self-organizing tendency. His concept of fallibilism assumes that error will eventually produce correction if inquiry continues long enough, almost as a logical property of the process rather than a result of anyone's deliberate effort.

In organizations, neither condition is reliably met. The inquiry is often not deliberate enough to surface semantic drift, and it does not continue long enough organically because people move on, priorities change, and no one is institutionally responsible for asking whether a term still means what it was originally defined to mean. Peirce's model assumes a community with both the patience and the incentive to keep correcting. Most enterprise data governance does not have either.

Wittgenstein: Meaning Is Use

Ludwig Wittgenstein held that the meaning of a word is, in many cases, best understood as its use in the language. Words acquire meaning through use in particular language activities, which he called language gamesRule-governed activities in which words have meaning through their use within specific contexts, rather than through fixed definitions.. Meaning cannot be fixed in advance; definitions do not govern their own application.

The enterprise implication is direct: a term defined in a governance document does not mean what the document says; it means what people do with it.

A system I worked on had the concept of a partner, originally meaning an external organization with whom we had a business relationship; however, since the infrastructure was already designed around this concept, when the system needed to support owned and operated properties, the term "partner" was carried forward for that purpose as well. This is not a partner in the original sense. The consequences in this case were minor; however, the drift went unnoticed precisely because there was no community of inquiry actively testing whether "partner" still meant what it was defined to mean. The same mechanism at scale, in a domain where the term governs something consequential, is where it becomes a serious problem.

Wittgenstein is describing how meaning works, not prescribing how it should work. He is not saying meaning should evolve through use; he is saying that is what meaning is. Language games shift, forms of lifeThe shared practices, activities, and cultural forms that give language its meaning within a community. change, and meaning drifts not because anyone decided it should but because the practice around the term changed and the term followed.

Semantics is therefore intended at modeling time, where we aim for certain meanings, but tested in use, where we discover whether those meanings actually hold up in practice.

Dietz: Essential Modeling and Transaction-Grounded Meaning

We do not need to model everything. The discipline Dietz applies in Enterprise Ontology is to model only what is essential to what the organization is and does. A schema created this way changes only when the organization's essential commitments change, not whenever observed reality across the organization shifts.

In DEMO, what anchors a concept is the product kindThe ontological fact that comes into existence when a transaction completes, serving as the stable anchor for a concept. of a transaction: the ontological fact that comes into existence when the transaction completes. "Sale 1087 is completed" is the stable anchor, not the transaction process itself. The process can vary in implementation; the product kind is what makes the concept precise and stable. Concepts defined this way are stable not because a committee agreed on them but because they are anchored to what the organization actually does. When the transaction changes, the concept changes with it, deliberately and visibly, rather than drifting unnoticed through informal usage.

Dietz also models how meaning travels between actors, extending the semiotic ladder into a communication architecture. The key point is that what actually travels between people is only the physical signal; everything above the medium level is reconstructed independently in the receiver's mind. Successful communication requires that both parties share the conceptual framework, the types, the language, and the social conventions that make reconstruction possible. If any layer is misaligned, the commitment the sender intended is not the commitment the receiver evokes, regardless of whether the signal arrived intact.

Communication Across the Semiotic Ladder

Sender Receiver Social / performa Social / performa Pragmatic Pragmatic Semantic Semantic Syntactic Syntactic Empiric Empiric Physical medium Physical medium Only the physical signal travels Meaning, intent, and commitment are reconstructed forma informa performa Communication succeeds only if the receiver can reconstruct the higher levels from the transmitted signal.

This diagram is a working simplification. The main point is that communication does not transfer meaning directly; it transfers a physical signal from which meaning, intent, and commitment must be reconstructed.

For DEMO, this means that the forma and medium levels are implementation concerns. The informa and performa levels constitute the essence of communication and therefore the essence of organizational modeling. This is the theoretical justification for why DEMO models only the O-organization and treats the I- and D-organizations as supporting infrastructure.

Information and Knowledge

If semantics is meaning, then information is structured meaning that can be communicated, stored, and reasoned over, but that lacks the pragmatic dimension of purpose.

Consider a birthdate record in a fintech system. The value is present, the label is recognized, and the semantic content is clear; however, the same record might be used to block a credit card application because the applicant is under 18, or to verify that a teenager qualifies for a supervised debit product. The semantics are the same; what changes is the purpose the system is serving and the obligations that follow from it. Axioms, constraints, and business rules provide structure and order to meaning, and they can be evaluated or enforced, but they still do not supply purpose. A rule can be correct and executable and still not count as knowledge.

The most consequential organizational knowledge, in Dietz's Enterprise Ontology, lives at the performa level, where social commitments bring facts into existence. My interpretation is that this is his answer to why semantics alone does not guide action. The practical question of how to establish that structure before deploying AI systems is addressed in Organizational Analysis Should Precede AI System Design.

The Schema Is Not the Meaning

The definition I am using here is that a conceptual schema is a structured specification of conceptual distinctions that supports shared reference, reasoning, and coordinated action. Its validity shows up in use, correction, and context, not in the specification itself.

A schema that is never queried against, never enforced, and never used to govern a decision may still exist as a specification; however, it is hard to say that it means much in practice. In roughly Peirce's sense, if a concept has no practical consequences, then its meaning is at best incomplete.

Conclusion

The problems described at the start of this note are not technical failures but are failures of semantic practice.

The semiotic ladder shows that most systems stop at the syntactic level, and stopping there does not make them semantic. This is as true for AI agents as for any other system; sophistication at the syntactic level does not cross the boundary into meaning.

What follows, at least in my view, is that we should model what is essential, ground meaning in transactions, make intent explicit, and keep accountability human.

Open Questions

On the interpretant for an AI agent. A language model does appear to provide an interpretant in Peirce's sense. Processing a sign causes an action in the model itself, the generation of a response, and that response can in turn cause a further response in whatever system or person called the model. Both conditions Peirce identifies are present: the sign produces a consequence, the generated output, and that consequence can itself become a sign that produces further consequences. Whether this constitutes a full triadic sign relation or a partial one, and what follows from the answer for AI governance, is a question this series will return to.

References