The Premise: Artificial intelligence is often portrayed as a technology of expansion - expanding knowledge, creativity, human capabilities, and even collective intelligence. Yet its most profound effect inside organizations may be quite different. By encouraging alignment, synthesis, and convergence in reasoning, AI could gradually reduce the cognitive diversity that human systems need in order to adapt. The risk may not be that machines think for us, but that they help make our organizations more conformist at precisely the moment they need greater divergence.
🪶🪶🪶
Can a technology designed to expand intelligence end up narrowing collective judgment?
For the first wave of digital transformation, the promise was straightforward. Information would flow more freely. Organizational silos would disappear. Decisions would become faster, more data-driven, and ultimately more effective.
Two decades later, artificial intelligence is accompanied by a remarkably similar narrative.
Yet something fundamentally different is happening.
AI does not simply accelerate the circulation of information. It increasingly shapes the conditions under which judgment is formed. It influences how problems are framed, how priorities are ranked, how alternatives are generated, and how solutions are evaluated.
In other words, AI is beginning to operate upstream of decision-making itself.
This distinction matters.
For decades, organizational power was primarily exercised through decisions. Today, it is gradually extending into the cognitive frameworks through which decisions become conceivable in the first place.
The shift remains subtle, but a growing number of signals suggest that it is already underway.
A recent BetterUp survey found that 54% of managers report receiving what has become known as “AI workslop” — content that appears polished, coherent, and professional, yet lacks genuine insight or discernment.
The problem with workslop is not that it is wrong.
The problem is that it is plausible.
The report looks credible.
The summary feels useful.
The recommendation sounds reasonable.
And yet, once consumed, little understanding has actually been generated.
What appears at first to be a quality issue may in fact reveal something deeper: a growing confusion between the production of content and the production of judgment.
Workslop, however, is probably only the visible symptom of a broader transformation.
The more consequential question lies elsewhere.
Across organizations, AI is increasingly used to clarify, summarize, structure, synthesize, and align.
These verbs appear neutral.
They are not.
Summarizing means prioritizing.
Structuring means imposing a logic.
Clarifying often means reducing ambiguity.
Aligning frequently means reducing divergence.
Every cognitive operation contains an implicit worldview.
Every summary elevates certain information while marginalizing other possibilities.
Every reformulation subtly alters meaning.
The question therefore becomes less technological than political.
What happens when a growing share of an organization’s intellectual activity passes through the same cognitive systems?
Consider a simple example.
A company begins experiencing unusually high employee turnover.
Before AI, several interpretations might emerge simultaneously. Some leaders would view the issue as a compensation problem. Others might point to a crisis of purpose, a leadership challenge, shifting cultural expectations, or changing generational dynamics.
These competing interpretations could create friction.
But that friction served a purpose.
It forced the organization to explore multiple representations of reality.
Today, the first instinct is often to ask:
“Analyze the causes of employee turnover.”
The machine does not make the decision.
Yet it immediately begins structuring the terrain on which the discussion will take place.
It selects variables.
It prioritizes causal relationships.
It organizes a narrative.
Long before the debate begins, it implicitly shapes what deserves attention and what does not.
This is where one of the most significant shifts may be occurring.
Power no longer resides solely in the final decision.
It increasingly resides in the way reality itself is rendered intelligible.
The implications may be more far-reaching than they first appear.
For decades, organizations have treated consensus as a sign of maturity.
Good managers create alignment.
Good leaders bring people together.
Good governance reduces conflict.
This logic has deeply shaped modern organizational culture.
Yet in an environment defined by volatility, uncertainty, and accelerating change, the equation becomes less obvious.
Human systems do not evolve through cohesion alone.
They also evolve through tension.
Major transformations rarely emerge from consensus.
They often emerge from persistent disagreement.
Weak signals almost always appear at the margins of dominant assumptions.
History repeatedly shows that minority voices sometimes see what the majority cannot.
The difficulty is that such voices rarely appear reasonable when they first emerge.
Recent history offers a striking example.
In the lead-up to the 2003 invasion of Iraq, several intelligence analysts in both the United States and Europe expressed reservations about the evidence surrounding weapons of mass destruction. The available information was ambiguous, and interpretations differed significantly. Yet as a dominant narrative consolidated, alternative assessments gradually lost visibility and influence. Looking back, the issue appears not to have been a lack of information but a declining capacity to assign sufficient weight to interpretations that challenged the prevailing assumption.
The data existed.
What diminished was the diversity of interpretations.
AI, by its very nature, tends toward a different dynamic.
It produces statistically probable responses.
It favors socially acceptable formulations.
It smooths rough edges.
It reduces extremes.
It brings positions closer together.
This tendency is not ideological.
It emerges from the architecture itself: identifying and reproducing dominant patterns across vast bodies of data.
In that sense, AI naturally promotes convergence.
This search for convergence is not new.
For several decades, organizations have increasingly rewarded those capable of smoothing interactions, reducing tensions, and maintaining cohesion. In my recent book The Power of Connection (FR version only), I describe this dynamic through what I call Level 3 of the Relational Evolution Scale (RES): the stage at which leaders become, above all else, architects of consensus.
For a long time, this posture was a strength.
In relatively stable environments, it encourages cooperation, reduces friction, and facilitates execution. Yet when a system enters a period of profound transformation, the very same pursuit of harmony can gradually become a limitation. When every rough edge is smoothed away, the early warning signals of change are often silenced as well.
Convergence is undoubtedly a powerful asset when it comes to synthesizing information. But it can become a weakness when preserving cognitive diversity. Adaptation rarely depends on a system’s ability to reproduce dominant patterns of thinking. It depends on its ability to detect what still lies beyond them.
A paradox begins to emerge.
Information abundance can conceal a scarcity of divergence.
Organizations may produce more reports, more analyses, more scenarios, and more ideas while simultaneously exploring an increasingly narrow cognitive landscape.
The quantity grows.
The apparent variety grows as well.
Yet beneath that abundance, cognitive diversity may be quietly shrinking.
As the same systems are repeatedly used to explore the same questions, familiar patterns of reasoning reappear, the same blind spots are reproduced, and similar solutions begin to dominate. It is no longer only the answers that converge.
The intellectual pathways themselves become increasingly alike.
This raises a question that remains largely absent from current debates about artificial intelligence.
Do organizations still need disagreement?
The answer may seem obvious.
Yet most contemporary systems are designed to reduce it.
Disagreement slows things down.
It complicates decision-making.
It generates uncertainty.
It disrupts coordination.
And yet it remains one of the most important drivers of adaptation, both individual and collective.
In relatively stable environments, consensus is often highly effective.
In rapidly changing environments, however, divergence becomes a strategic resource.
It allows organizations to explore alternative futures.
It reveals blind spots.
Most importantly, it protects systems from becoming trapped by their own assumptions.
Natural ecosystems offer a useful analogy.
Their resilience does not emerge from uniformity but from diversity.
A seemingly marginal species can suddenly become essential when environmental conditions shift. What appeared unnecessary yesterday may become critical tomorrow.
Biodiversity functions as a reservoir of adaptive capacity in the face of uncertainty.
Human systems appear to follow a similar logic.
Minority opinions, dissenting intuitions, and unconventional interpretations often play the same role. They may seem inefficient, uncomfortable, or excessively cautious—until a change in context suddenly reveals their value.
We may be entering a period in which cognitive diversity becomes as strategically important as cybersecurity, data governance, or technological infrastructure.
Organizations may need to learn how to protect spaces where minority viewpoints can exist.
They may need to preserve zones where uncertainty is not immediately eliminated.
They may need to accept that certain tensions are productive and that everything which slows a system down is not necessarily inefficient.
Some forms of friction are, in fact, the very conditions that make emergence possible.
Artificial intelligence will undoubtedly expand many human capabilities.
At the same time, it may reinforce a tendency already deeply embedded within modern organizations: the relentless pursuit of harmony, alignment, and coherence.
The question, then, may not be whether AI will make us more intelligent.
The question is whether, at the precise moment human systems most need cognitive plurality, they will still be capable of preserving disagreement, minority intuitions, and the relational tensions that make adaptation possible.
An organization capable of producing consensus is often an effective organization.
It streamlines decisions.
It reduces internal friction.
It improves coordination.
Yet the major economic, political, and institutional disruptions of history remind us of a more uncomfortable reality.
Human systems rarely fail because they lack cohesion.
They fail because they continue responding to a world that no longer exists.
What enables adaptation is not alignment alone.
It is the capacity to make room for what disturbs, challenges, and contradicts prevailing assumptions: weak signals, dissenting perspectives, and alternative readings of reality.
Divergence is not a malfunction of the system.
It is often the way the future first begins to reveal itself.
An organization capable of preserving cognitive diversity protects far more than freedom of expression.
It protects its capacity to evolve.
Consensus helps organizations optimize the present.
Divergence helps them adapt to the future.
About Human Dynamics Group
Human Dynamics Group explores how contemporary technological, cognitive, organizational and geopolitical transformations are reshaping human systems.
Our work focuses on:
cooperation dynamics ;
leadership transformation ;
regulation mechanisms ;
collective tensions ;
discernment capabilities ;
and the new forms of human organization emerging under acceleration.
Through publications, research, analyses, conversations and strategic interventions, we seek to better understand what weakens — or strengthens — human capacities to:
perceive,
cooperate,
decide,
and sustain collective cohesion in increasingly complex environments.
→ To follow our work or connect with us: Here