Rethinking the enterprise

Why hierarchy is a natural dissipator of knowledge

Each level of the organisation chart filters, distorts and slows down the flow of information. The result? Decisions made on a fraction of reality

by Emiliano Pecis*

16 January 2026

4' min read

Translated by AI

Versione italiana

4' min read

Translated by AI

Versione italiana

In the Rethinking the Enterprise column, we often ask ourselves when management, as we know it, is serving its purpose. An interesting perspective is to understand how freely information circulates in the hierarchical organisation chart to give top management the data it needs to make decisions. Our analyses always start with the best case scenario, because as Deming liked to repeat, 94% of the problems in a company are caused by the system, not the people.

We have already defined the hierarchy as an organisational anxiolytic: this is because it promises clarity and coordination but rarely guarantees them. In practice, the more information flows through hierarchical levels, the more it is distorted by omissions and reinterpretations. Paradoxically, management functions like a cordless telephone, producing plausible but often distant versions of the facts.

Five filters that distort reality

Every organisation functions as an imperfect information circuit, traversed by five structural filters.

The first is cognitive: no one really understands the entire domain of what they handle. The second is interpretive: as Karl Weick recalled, people do not convey information, but construct meaning, translating facts into narratives consistent with their experience. The third is role-based: as James March noted, each individual acts according to what is 'expected of someone like him', remaining consistent with his role even when he knows better.

To these are added two further filters: the political filter, because each manager protects his own area and manipulates, even unconsciously, what he communicates; and the cultural filter, because companies reject what contradicts their internal narrative. These five filters operate in both directions: they deform both what goes down (strategies, directives, priorities) and what comes up (reports, feedback, alerts). The result is a closed loop of altered, self-confirming and self-asserting knowledge.

Information Decay

Looking mathematically (**) at the flow of knowledge within a hierarchy, we discover an exponential decay: at each level a fraction of truth is lost. Even under near-perfect conditions - competent people, transparent communication, fast turnaround times - the amount of correct and useful information is significantly reduced. And when delays, reporting times and co-ordination costs are added, the actual value of the information reaching the top never exceeds half of its original value. In the best case scenario, information that starts at 90% accuracy may reach the operational team with 60-70% fidelity, and return back to top management with less than 30%. Not because someone has manipulated it, but because each step - each cognitive, interpretative, role, political and cultural filter - has eroded a piece of reality.

An invented but plausible example: the case of the North-East

In a manufacturing company, sales in the North-East start to decline. The agents on the ground, who talk to customers every day, immediately realise what is wrong: a local competitor delivers faster, the new after-sales service is slow, and the updated price list has made entry-level products less competitive. The real causes are clear and concrete. But reality begins to warp as soon as information enters the chain of command.

The area manager knows that the marketing management strongly wanted that new price list and that questioning it could be read as a political attack. So, in his report, he speaks generically of 'seasonal downturn' and 'slight contraction in demand'. Role filter and political filter: he communicates what is compatible with his role, not what he knows.

The document reaches the national sales management, where another distortion takes place. The news is reinterpreted as 'lower propensity to buy of SMEs', a more reassuring and easier macroeconomic phenomenon. Interpretation filter: facts are adapted to an already known mental scheme.

The CEO sums it all up: 'Demand is slowing down in the north-east due to the uncertain economic climate'. Cultural filter: the company is convinced that it is 'strong on price and reliability' and finds it hard to conceive that therein lies the problem. The final message sounds plausible, coherent, perfectly logical. But completely false.

When the information descends again - after meetings, delays and presentations - it encounters the last cognitive filter: management, now far removed from operational detail, no longer has the tools to validate the diagnosis. The CEO reacts as rationally as possible to the picture in front of him: he approves a national discount and advertising campaign to 'boost demand'.

But the problem, on the ground, remains the same. The price list is wrong, the after-sales service is slow, and no campaign can change that. Information has degraded like an experiment in management entropy: from a concrete signal to a macroeconomic narrative. The summit decided with a tenth of the initial truth. And the real truth never made it to the top.

The structural limit of the hierarchy

Even the most efficient company cannot evade this law. Every hierarchical step is a loss: knowledge is consumed, distorted, forgotten. And since information costs money, in time, in meetings, in mediations, what reaches the top is always a filtered version of the real thing. Hierarchy, in short, is a dissipative system of knowledge: it turns truths into interpretations, interpretations into decisions, decisions into new problems to be interpreted.

A possible role for AI to reduce filters

We need to shorten the information chain, create direct channels between those who know and those who decide, restore autonomy to those who operate close to reality. The technologies of artificial intelligence can help, but only if they are designed to illuminate flows, not control them.

An inference system can link directly to sources of truth and return a concise but verifiable picture to the top, unmediated by roles or narratives. The real challenge is no longer to have information, but to preserve its quality along the way.

As long as knowledge has to cross five floors and three committees to get to the top, management will continue to make decisions with a fraction of reality. And as long as the truth has to ask permission to rise, no organisation can truly improve.

(**)| The mathematics of hierarchical dispersion

I hypothesised a formula to model this phenomenon: U = p₀ × ∏(κᵢ × θᵢ × φᵢ), where κ measures how much is understood, θ how much is communicated, φ how much remains true. Even under optimistic conditions (κ=0.95, θ=0.90, φ=0.92), each level retains only 79% of the information. After three levels, 49% remains. In a complete round trip through 3 levels (6 steps): less than 38%.

*AI & Organisational Strategist