Ask a founder of a forty-person company to name one person whose departure would stall a specific capability inside of a week, and almost all of them can. They can name the person, name the capability, and describe exactly what would stop moving. What most of them have not done is write that name down and count how many names are on the list.
The departure is when the unprepared pay for it
The departure is the moment a concentrated capability becomes visible, but the risk has been present every day before it. A capability that lives in one person and nowhere else is a standing condition, present the whole time, not only on the day the person leaves. The organization pays on the departure date for an exposure that has been true for years.
Framing concentration as a departure problem is what keeps it from being addressed. A departure feels like a low-probability event, and the standing condition feels like normal operation, so the two get separated and the work of addressing the standing condition never gets scheduled. The risk was always there. The departure is the billing date.
A capable person looks the same as a healthy organization
The expert being good at the job looks identical to the organization being healthy. The work gets done, on time, to standard, by the person who has always done it. Nothing flags the concentration. The reports show the work moving. The customer sees the output. The founder sees the person performing. The concentration is invisible from every angle that measures output, because output is exactly what the expert is producing.
This is the confusion that keeps the condition in place. Having a highly skilled employee is good. A skill that exists nowhere else and in no recorded form is the risk. Those two states look the same from the outside, and only one of them is a risk.
Cross-training is the move that feels like a fix and is not. Pairing a second person with the expert helps, but without actual documentation the knowledge island simply becomes a knowledge island of two. The second person depends on the first being present to learn. The capability still has no written form that survives both of them leaving, and the organization still cannot point to where the skill lives except in the people who hold it.
The capability disappears immediately, not gradually
A concentrated capability does not degrade when the person who holds it leaves. It disappears outright. The organization discovers this under pressure, mid-task, rather than at a convenient time to plan around it, because the gap was never written down and the people who would absorb it were never identified.
Take a forty-person commercial AV system contractor where one project manager has, over nine years, become the only person who knows which of the company's fifteen general contractor clients will accept a substitution on a spec when the listed unit is backordered, which ones will push back on a change order, and which ones need the paperwork notarized a particular way. The substitutions, change orders, and notarization rules have never been written down. When she takes a week off, three jobs slow, and the people covering improvise. The two weeks of vacation each summer were the only time anyone else touched the function, and even then the hard cases waited for her return. When she retires, the company loses a capability it took nine years to build, in an afternoon, with no record to train a replacement from.
Concentration also caps growth, quietly
A capability concentrated in one person has a second cost that shows up before any departure. It cannot scale beyond that person's available hours. The ceiling on the function is the ceiling on one person's week, and the organization reads the ceiling as a market or process limit rather than what it actually is: a single-person dependency that no amount of demand will move.
A founder who watches a specific function hit the same wall every quarter, no matter how much capacity is added around it, is often looking at a knowledge island misread as a throughput problem. Adding headcount elsewhere does not raise the ceiling, because the constraint is not labor in general. It is the absence of a documented form of the capability that a second person could perform from. The demand is there. The growth is there. The capability is not, because it has never been written down in a shape that more than one person can hold.
The inventory comes before the fix
The first move is an inventory. Which capabilities are concentrated in whom, and what is the exposure if that person is out for a week versus out permanently. Documentation projects and transfer programs come after, and both tend to stall when the list of islands is not known, because pointing a documentation effort at an unnamed set of capabilities produces a pile of write-ups no one has prioritized.
The inventory is short and uncomfortable. Most owners can produce a first draft in an afternoon, from memory, because the islands are visible to the owner and hidden from everyone else, and from the organization's records. The owner already knows where each one sits and who holds it. The first pass does not have to be complete. It has to be honest, and it has to separate the islands that would stall a function for a week from the ones that would stall it permanently. The exercise is writing down what the owner already knows and has not yet counted.
Concentration is a standing condition the owner is carrying
The point is to read concentration as a condition the owner is already paying for, in risk, every day, not only on the day of a departure. Naming the islands is the move that every later step depends on. Documentation, cross-training, transfer, and automation all require a capability to be identified and located before any of them can be pointed at it. The owner who has the inventory can decide which islands to address first and which to accept as a known exposure. The owner who does not has the same exposure and no way to prioritize it.
The islands are hidden from everyone but the founder, and that is exactly why they stay concentrated until the day they are not. The inventory turns an unnamed exposure into a named one, and a named one is the only kind that can be addressed on purpose.