Are organizations losing their distinctiveness through AI?

What has become the norm on LinkedIn describes the symptom: texts that are polished, clear, to the point, and at the same time so lacking in individual stylistic imprint that they could easily be swapped. Between people, companies, industries, and cultures. Cultural nuances disappear, the very cues that used to signal who is speaking and from what context.

What concerns me here goes beyond style. AI detects patterns and reproduces them. That makes texts more readable and more scalable. What it does not improve, and may even actively weaken, is an organization’s ability to generate and preserve a distinctive stance of its own. When the tool used for communication brings its own cognitive logic with it, clarity through contrast, meaning through groups of three, conclusions always moving toward a solution, then over time this logic shapes not only external communication, but also the internal structures in which thoughts are formed.

Initial research is now providing evidence for what I am describing here. A study presented at the CHI Conference 2025 found that participants who had temporarily worked with AI-generated strategies subsequently produced significantly more similar ideas than before, even without AI. The homogenizing effect does not stop once the tool is put aside. It remains. A recent paper in Trends in Cognitive Sciences captures this precisely: “The Homogenizing Effect of Large Language Models on Human Expression and Thought.” When patterns of argumentation, prioritization, and decision-making are shaped extensively by AI, it is not only outward-facing language that changes. What shifts is what internally counts as a good thought.

That is why I keep asking myself: are organizations losing their distinctiveness through AI?

By distinctiveness, I mean something deeper here than the tone of a press release: the ability to frame problems in a way that emerges from a specific history, a specific perspective, and a specific stance.

The erosion of this ability is hard to notice because it does not arrive as a decision. An organization does not decide to become generic. It loses its distinctiveness because the questions it asks internally shift almost imperceptibly, toward what a language model considers a sensible question. Briefings become more precise, but less surprising. Strategy papers show that someone has worked thoroughly, but no longer that someone brings a particular point of view that no one else would formulate in quite the same way. What disappears are precisely these movements of thought. And they are often the decisive ones.

This has consequences that go beyond communication. Differentiation is the result of a consistency of decisions that would not be self-evident to others: which customers an organization rejects, which risks it takes, how it responds to crises. Positioning documents do not create that. This consistency draws on a stance that, at best, runs through the entire organization, from strategy to day-to-day communication. If the tool that shapes communication systematically smooths out that stance, the organization loses more than its voice. The very foundation from which decisions become recognizable and comprehensible begins to erode.

Those who sound like everyone else eventually start thinking like everyone else. And then they start deciding like everyone else, too.

The obvious counter-question is: can we not train this? And the answer is: partly, yes. Surface-level linguistic features, vocabulary, sentence rhythm, tone of voice, can be taught to a model. That works, and many companies are already doing it: with custom GPTs, fine-tuning on internal documents, and detailed style guides used as prompt context.

What cannot be trained is judgment. A model learns what has been expressed in past texts. Yet what truly makes an organization distinctive is often what has not been expressed: the question that was asked against the obvious, the decision that looked wrong on paper and nevertheless turned out to be right, the instinct that appears in no strategy document. These things do not exist as training text.

And then there is a fundamental limitation of the current paradigm. Language models tend toward the middle: they predict what is statistically most likely. Fine-tuning can shift the mean, but the principle remains the same. The more unusual a situation is, the more strongly the model pulls back toward the statistical middle. Distinctiveness, however, emerges precisely in unusual situations, when an organization’s own answer is not the most probable one.

Here, the tool reaches a limit that cannot be overcome through better prompts. What creates distinctiveness is the opposite: the deliberate deviation from what is probable. This shifts the strategic task of communication work. The primary challenge is to make sure the frame is right before the tool gets involved. Whose thinking is embedded in the briefing? Whose stance carries the argument?

That requires people who know the obvious path and deliberately move against it. People who know which questions an organization has not yet asked, and ask precisely those. People who know the organizational history that is not written down in documents. People who recognize when a situation is unusual enough that the most probable path must not be their own. That is a human capability, and it becomes more valuable the more everything else can be automated. It cannot be delegated, not to an agency, and certainly not to a language model.