How Organizations Actually Work
There are two organizations in every building. One is described in documents. The other gets the work out. Most attempts at change are addressed to the first one.
Two organizations, same building
The designed organization has a structure, a set of processes, defined ownership and a stated way that work moves. It exists in slides, handbooks, system configurations and the answers people give when a senior person asks how something works.
The operating organization is what happens. It has different paths, different decision points, and a different set of people who matter. It includes the message sent before the ticket, the person who is asked because they always know, the step that is skipped because it stopped being useful, and the spreadsheet that holds the version everyone trusts.
The second one is not a shadow or a corruption of the first. It is the organization. The first is a description of it that has fallen out of date, in places, in ways nobody has audited.
The divergence is not uniform, which is what makes it hard to spot. Some parts of the documented system are perfectly accurate and have been for years. Others were accurate at the time of writing and stopped being so after a reorganization, a system migration, or the departure of one person who had been holding a seam together. The document does not distinguish between the two, so reading it gives you a mix of fact and archaeology with no marking to say which is which.
The undesigned version is coherent
It is tempting to read the gap as decay. Discipline slipped, people cut corners, the process needs reinforcing. That reading is almost always wrong and it leads directly to the wrong intervention.
What the gap actually records is adaptation. Every deviation started as somebody solving a real problem the design did not handle. The skipped step was skipped because it produced nothing. The private record exists because the official one could not be acted on. The informal approval happens first because the formal one takes four days and the customer is waiting.
Which means the operating organization is a rational response to conditions, assembled locally, one decision at a time, by people with better information about their own work than anyone above them has. It is not indiscipline. It is a distributed repair of a design, carried out without authority, budget, or anyone noticing.
Coherent does not mean optimal, and the distinction matters before this becomes an argument for leaving things alone. Local repairs are made with local information. The person who built the workaround could see their own step and the two either side of it, not the whole flow, so they solved their problem in a way that may have moved cost onto someone three handoffs downstream who has since built their own repair on top of it. What accumulates is a structure where every individual piece has a good reason and the assembly has none.
The repairs also outlive their causes. A check added after an incident in 2019 is still being performed in 2026 by people who have never heard of the incident, because nothing in the organization is responsible for retiring a repair once the condition it addressed has gone. Accumulated systems only ever add.
Resistance is not people defending their comfort. It is people defending a repair, and they are the only ones who know what it was repairing.
Why it stays invisible from above
Nobody hides any of this. It stays invisible for structural reasons.
Reporting is generated by the designed system and describes the designed system. If the workaround happens in a message, in a spreadsheet, or in somebody's memory, it produces no data, so from a distance the organization appears to be executing its documented process. The absence of evidence is the design working as specified.
The people who could describe it have also stopped noticing. Ask someone how a request gets through and you will get the official answer, not because they are being careful but because the compensating is now automatic. It became part of the job years ago and it does not register as an event. I have sat with someone while they described a process, then watched them do something quite different an hour later, with no sense of contradiction between the two.
There is a third reason, and it is the least comfortable. In many organizations the workaround is technically a deviation from policy, which means describing it accurately to a senior visitor carries a small risk and no reward. Nobody lies. People simply answer the question that was asked, in the terms it was asked in, and let the rest go unmentioned. The more the organization has punished deviation in the past, the cleaner and less true the answers get.
What this does to change programs
An initiative aimed at the designed organization can succeed completely and leave the operating one untouched.
The process is redesigned, the system is reconfigured, the training happens, and the people doing the work map the new instructions onto whatever they were already doing, preserving the parts that were load bearing. Compliance is high. Behavior is unchanged. Everyone is being reasonable, and the organization has spent a year negotiating with a description of itself.
The failure mode has a recognizable signature afterwards. Usage numbers are healthy, the sponsor's report is green, and the operating measure that justified the whole thing has not moved. When that combination appears, the useful question is not what went wrong with the rollout. It is which repair the new design failed to account for, and where that repair went once the new system arrived, because it did not disappear. It moved.
The uncomfortable corollary is that resistance is frequently the most useful signal available. When people push back on a redesign, they are usually not defending their comfort. They are defending a repair, and they are the only ones who know what it was repairing.
Which suggests a cheap discipline: when someone objects to a change, treat the objection as a bug report about the design rather than as a position to be overcome. Ask what breaks, specifically, in a case they have actually seen. Most objections resolve into either a real omission, which you now have for free, or a preference, which becomes obvious under the same question. Either result is worth more than the persuasion campaign that usually gets run instead.
Reading the difference deliberately
If the operating organization is the real one, the practical question is how to see it, given that it does not describe itself and its inhabitants have stopped noticing it.
Three approaches recover most of it. Follow the work rather than the org chart, because the path a request takes crosses boundaries that no structure diagram predicts and reveals the people who matter regardless of where they sit. Look for artifacts rather than asking for descriptions: the shared file, the group chat, the saved view, the document called something like "how it really works" that exists in a folder nobody senior opens. And ask about the last difficult case rather than the general process, because a specific recent instance is remembered accurately while a general description is reconstructed to match the official version.
What none of that requires is authority or a mandate. It requires being present and being uninteresting enough that the work carries on normally while you watch, which is a real constraint on who can do it and one of the better arguments for the person doing the diagnosis being close to the operation rather than visiting it.
Every organization already has an operational system
The practical conclusion is not that documentation is worthless or that formal process should be abandoned. It is narrower and more useful than that.
Every organization is already running an operational system, made of people, workflows, decisions, data and technology, arranged in a particular way. The only variable is whether that arrangement was designed or accumulated. Accumulated arrangements are not random, but they were optimized locally, under old conditions, by people solving the problem in front of them, and they carry the cost of that forever.
Designing one means starting from the arrangement that exists rather than the one on the wall. That is a slower beginning and it is the only one that has a chance, because the operating organization is the one that will decide whether anything you build survives contact with the work.
If this describes an operation you are responsible for, the diagnostic is where that conversation starts.
How engagements begin
