Every operation has someone like this. You ask how a particular thing actually gets done, and the answer is not a document or a system. It is a name. “Oh, Priya handles that.” And for a long time, that works beautifully. Priya is quick, she knows every exception, she never drops it. The thing just gets done, and nobody has to think about how.
Then Priya goes on holiday. Or off sick. Or hands in her notice. And suddenly the thing that always just happened does not happen, and nobody left in the room knows why, or how, or what to do about it.
This is key person dependency, and it is one of the quietest and most expensive risks an operation can carry.
It is dangerous precisely because it does not look like a risk. It looks like competence. A capable person absorbing complexity so the rest of the business does not have to feel it is, day to day, a good thing. The problem is that every time they absorb another exception, another workaround, another “I will just sort it,” the knowledge moves further out of the system and further into their head. The operation keeps running, so no one notices it has quietly become dependent on a single person’s memory.

And that knowledge is almost never written down. It is not in a process document, because the person who holds it has never needed one. They are not hiding it. It simply lives in the one place nobody else can read: experience. Which decisions matter, which steps can be skipped, who to call when the system says no. None of it is on paper, because for the person who knows, it is just obvious.
The cost stays invisible until the moment it is enormous. While the person is there, the operation looks healthy. The day they are not, you discover how much of your business was actually running on them, and you discover it at the worst possible time, under pressure, with a customer waiting. The bill for years of undocumented dependency arrives all at once.
The fix is not to make your best people less capable. It is to get what is in their heads out into the open while they are still here to explain it. When I map an operation, I watch for the steps that resolve to a name rather than a process. Those names are the single points of failure. For each one, sit with the person, write down what they actually do, including the exceptions and the judgement calls, and build the smallest possible fallback so the task survives a week without them.
A good test is simple. Pick the person you would least like to lose, and ask what would break on the Monday morning after they left. If you cannot answer in specifics, you have not found your risk yet. You have just been lucky.
Finding the steps that depend on a person rather than a process, and getting them out of people’s heads, is a lot of what a Service Clarity Audit does.