Tribal Knowledge, and Antifragility

I recently spotted this tweet:

I've made the point before - it is less efficient, in a bottom-line, quarter-to-quarter profit margin way to maintain process knowledge of everything you use and depend on. Yet, you can't afford to not maintain that knowledge. This goes to computer manufacturing, steel manufacturing, and even something as mundane as customer service/support desks.

Take the recent "COBOL crisis."

It is true that, as Wired points out, the errors generally being seen are related to the front end webservers and more modern "robust" languages running on them, yet, digging through that article it is plain that there are still issues that need a lot more COBOL experts now or soon, to upgrade the systems, the hardware they run on, and so on, without bringing everything to a crashing halt.

In all fairness, I'd bet a lot of the affected organizations actually regularly update the hardware as needed, but it does point, again, to depending on systems that have a limited number of experts who can work on it, and not having people on tap to draw on when problems come up or the workload increases, gambling that it will come up rarely enough, or you can find contractors quickly enough, to make do.

The contractors, on the other hand, if you really need them, won't have the intimate knowledge of working with your systems day to day.