When the Machine speaks, who is talking?

Communication, in large part of the twentieth century, has been understood, taught and propagated as a ‘signal-transmission’ concept. This idea was proposed by Claude Shannon in 1948, where he presented the mathematical theory of communication. It became the most seductive and misleading metaphor of communication primarily because of the over simplification of the communication process that it offered. Communication, as he presented, was a signal transmission problem which involved the sender, channel, receiver, and had noise. Success meant the message arrives intact and therefore indicating that the process can be engineered.

Over a period of time, this view of communication came out as highly flawed and ironically, in 2026, we have developed a technology that perfects this model of communication and we have once again labelled this perfection as ‘communication’ and ‘intelligence’.

The communication argument that ran for a century

Communication debates over the 20th century put our every conceivable argument to contest Shannon’s understanding of communication hinged on the core thesis that this model simply captured the plumbing but missed the point entirely.

Communication is not just delivery, and neither is even an exchange. At its core, communication is the act of constructing a shared meaning and that act, every school of thought insists, is irreducibly human. Every school of thought, from the phenomenologists to the social constructivists, from Stuart Hall's encoding-decoding model to Watzlawick's axioms of human communication, was making a version of the same argument: the model captures the infrastructure and misses the act entirely.

The first objection was about the receiver that the model treated as a passive recipient. Communication evidences indicate that actual human beings receiving messages and making meaning of it is considerably more complicated. The same message, received by the same person in a different emotional state, produces a different meaning and the same message received by a different person with different stakes and different history, it can produce an entirely opposite meaning. The message does not travel like a packet. It arrives, and then it is made.

The second objection was more fundamental. A significant strand of communication scholarship argued that communication's primary function has nothing to do with information transfer.

People communicate to build relationships, to signal care, to create the experience of being understood. The words chosen are not merely vessels for content. They are the content. When a chairperson's note tells shareholders the firm navigated a difficult year with resilience, they are not reporting facts. They are constructing a version of reality, making deliberate choices about what to foreground and what to leave in the shadow, where to claim credit and where to quietly reassign blame. This is not writing. It is thinking made visible.

That distinction matters enormously. Because thinking, unlike writing, cannot be automated

What does AI and Large Language Models actually do?

Large language models (LLMs), that seem to be substituting for human communication, actually work on predictions. Trained over large sets of textual data, this technology has mastered in predicting the most statistically probable next word given everything that preceded it. This is done with such extraordinary sophistication, and the output generated is often fluent, structurally sound, and contextually responsive that it can be mistaken for communication. In the engineering sense, it is good communication. Clean signal. Low noise. Reliable delivery.

It is not, however, understanding, which is the core to communication. And the reason this distinction matters is not philosophical vanity. It matters because understanding, from the outside, looks almost exactly like sophisticated pattern completion. The model produces writing that resembles human writing because it has been trained on the most widely circulated human writing. It does not know what it is saying. It knows what tends to come next.

And here I do not mean to criticise technology but describe precisely how it works and to highlight the associated risks which not merely is misinformation, or hallucinations but the human misunderstanding. The risk is that we as humans mistake the fluency of the technology for something it does not possess, and in that mistake quietly surrender the only thing that distinguishes communication from a mere transfer.

The consequence of AI mediated communication: We all will sound the same

The greatest threat of AI-mediated communication is the high likelihood of homogenisation. With the growing number of users employing the same model to draft their communications, what they are producing is not a million distinct messages but variations on a probability distribution.

The model has learned from the most widely read human writing and produces writing that resembles it, which means balanced, cautious, structurally conventional, and smoothed toward the statistical average. Corporate English is already the most relentlessly averaged form of language in circulation. AI does not correct this tendency. It scales it; industrialises it and it is concerning because uniqueness and authenticity in the communicative voice is not an aesthetic preference.

It is through communication that organisations express their identity and discover what they actually believe in. The email that flouts the template, the chairperson’s letter that speaks the difficulties of a genuinely difficult year, the proposal that argues against the consensus are often the most consequential communications a company ever produces. They require someone willing to make a choice and place their own judgment visibly on the line that the probability engine would not make.

Watzlawick's first axiom of communication states that one cannot not communicate. Every silence carries meaning. Every evasion is a statement. Every word chosen over another word is a choice with consequences. A world of AI-generated communication is a world in which those choices are quietly outsourced to a system that cannot, by definition, have intentions. The signal keeps arriving. The meaning slowly drains away.

The human intelligence that AI is still far from

While the LLMs have processed text at a scale that no human mind can approach but it does not have the intelligence or knowledge that accrues from experience which is strictly human. The experience of from having been wrong in front of people who trusted you, from watching a difficult conversation deteriorate and knowing precisely why; from caring, genuinely and personally, about the person on the other side of the message.

A communicator who thinks does something specific. They re-create the situation, realities, emotions and perspectives of the audience before they ask for facts, build arguments and finally draft the communication. The communicator creates a world of the audience to think through what the audience most needs to understand, what the audience will most resist, and what they themselves are willing to be honest about. They know that every communication is a small act of world-building, and they take the responsibility that comes with it. Therefore, they may sometimes communicate through silence.

This is what a century of communication theory was arguing for. Communication is not, at its root, a technical act. It is a moral one. It involves a mind that chooses, in full awareness of consequences, what meaning to create and what reality to construct in another person's understanding. That is not a function. It is a judgment. And judgment is exactly what the pattern-completion engine cannot exercise.

Prudent choices for the future

There is a version of the AI-communication future that works. The machine handles the mechanical labour of drafting while the human thinks more carefully, with more time and less friction, about what actually needs to be said. Judgment deepens because effort is redirected. The tool serves the thinker. This version is available to us.

To conclude, I would therefore, I would say that , use of LLM and AI-mediated communication is not a technology question. It never was. It is a thinking question. And thinking, as it turns out, was always the skill. We just forgot to say so, right up until the moment a machine offered to say it for us.

(Dr Ruchi Tewari is the Associate Professor and Associate Dean – Marketing, Communication and Public Affairs, Chief Marketing Officer at MICA. An academician with over 25 years of experience, she holds a PhD in Management with a focus on CSR communication in the Indian IT sector. She also holds an MPhil and an MA in English Literature.)