For all its flaws, Wikipedia is one of the marvels of the modern world. Free access to encyclopedic information about pretty much any subject you can think of. Maintained by thousands of international volunteers around the clock. Everything tends to be categorized, organized and kept relatively up-to-date. You can start at something as simple as fish and with a few clicks end up at the highly specific Eucarida.

This model for organizing information works decently well. Wikipedia is perhaps the best known, but there are also thousands of very specific wikis which dive even deeper into a subject. For example: the Harry Potter fandom wiki or Glottopedia for linguistics. Somehow, this model also became very popular within enterprises. Companies using Jira often have an internal Confluence. There are also less common alternatives such as a standalone MediaWiki instance or Gitlab Wiki.

Your internal wiki isn't Wikipedia

But here's the catch: your internal company wiki isn't Wikipedia or that of a highly opinionated self-organizing fanbase such as Harry Potter. Most company wikis start off really well. "Everyone can contribute! Just put it on the wiki!". But before you know it spaces and pages get created left and right. The idea of ownership and governance doesn't exist. "Everyone's responsible!". In practice that means nobody is.

Once the initial enthusiasm fades, the cracks start to show. You need to be "in the know" of which pages are useful. Bookmarks to useful pages are shared like ancient relics. Nobody dares to delete anything - what if it's important later? Who even owns this information? Slowly but surely the wiki fills up with outdated, irrelevant or/and incomplete information.

At this point, searching through the wiki becomes a running joke. If you are not "in the know" of which pages are good, you are looking for a needle in a haystack. You might even lack permissions to access certain content due to different teams or departments organizing things separately because they want to silo themselves from the current state of affairs.

Disdain against the wiki might start to grow. People are frustrated that it’s unorganized and that found information cannot be trusted. Or can. But who knows at this point? Yet nobody wants to take responsibility to clean it up. After all, who has time to sift through sometimes hundreds of pages of outdated content? Creating a new page (the-current-year-version-which-is-correct) is much easier. The wiki becomes a tool people avoid, even though it was supposed to make their lives easier.

Information debt and information bankruptcy

Similar to technical debt, information debt starts compounding when you let outdated, redundant or incomplete information grow wild. Before you know it the effort required to clean it up can't even be measured anymore¹. You have then reached information bankruptcy - existing information becomes so unmanageable that it’s effectively useless.

Sadly, I don’t have a cut-and-dry solution for you right now. The first step is admitting that there is indeed a problem. From there, start with the hard part: removing everything that no longer adds value. This will feel uncomfortable, and you will likely face resistance, especially from the "just-in-case" crowd who want to keep it in case it is ever needed.

One suggestion I do have is to ask the hard question: does anyone actually need this information? If someone is willing to take ownership of it, great! But if nobody steps up, it’s time to let go. After all, if no one cares enough to claim it, why keep it? Ownership shouldn't be temporary either. Otherwise the information will just get outdated or redundant over time which results in even more information debt.

There is no foolproof strategy to handle this kind of problem. It's unlikely your internal company wiki will get the thousands of volunteers Wikipedia has. But you will definitely need at least one (or a few) to systematically keep things in check because no matter how small or big the organization, information (and the wiki) doesn't just organize itself.

1. No matter what the marketing & shiny AI buttons tell you, loading all your information debt in an LLM doesn't solve this. Information in ≈ information out.