The Myth of Relational Knowledge

AI didn’t change what good community knowledge looks like. It just made bad architecture *really* noticeable.

There are three kinds of community managers right now.

The ones who are quietly dismantling their communities because AI is going to make them obsolete anyway.

The ones who are carrying on like it’s 2019 and nothing fundamental has shifted.

And then there’s a smaller, quieter group. We’re a mix of deeply understanding what’s happening and adapting accordingly… or those just getting on with it because why the heck not?

This article is for that third group. But the other two are welcome to stay. Maybe we’ll convert you to the dark side.

But first, we’re going to need you to accept a really harsh reality: AI didn’t change what good community knowledge looks like. It just made it impossible to pretend that bad knowledge architecture was fine.

Do you know the move I’m talking about? Someone posts a question your community has definitely answered before, and instead of pointing to a crisp, documented resource, someone types: “Oh, great question, you should ask Jillian, she handled this exact thing last quarter.” And Jillian answers, because Jillian is generous and competent and has the institutional memory of a small government agency. Crisis averted. Thread resolved. Nobody notices that the actual answer lives entirely inside Jillian’s head.

That worked. Until it didn’t.

The communities that are panicking about AI right now were always one Jillian-on-PTO away from collapse.

Think about what “ask Jillian” actually means. It means your knowledge was never really in your community. It was in a person, and that person happened to be reachable. The member who joined six months ago and doesn’t know Jillian yet? They either figured it out themselves, gave up, or got lucky. That’s not community knowledge. That’s a single point of failure with good vibes.

And yes yes - we know the retort: “But but but… AI!”

AI didn’t create this problem. It removed the social scaffolding that let us ignore it.

Here’s where I’d gently push back on a story we tell ourselves about human cognition: we like to think we navigate knowledge relationally, that we’re fundamentally different from machines because we connect ideas through people and context and lived experience. And sure, there’s something to that. But mostly? We know where things are because we have context. We remember that the onboarding doc lives in the folder that so-and-so reorganized last spring. We know the workaround for that integration bug because we were in the Slack thread when it happened. That’s context. Not magic. Not some ineffable (how great is that word?) human quality that machines can never touch.

The gap between how humans find knowledge and how AI finds knowledge is narrower than we’d like to admit. (I know, I know. Uncomfortable. Stay with me.)

Which means the communities that will hold up, the ones that serve members well regardless of whether those members arrive having already asked ChatGPT something, run it through their company’s AI tool, or just typed a search query like it’s 2011, are the ones that were already doing knowledge architecture right. Clear, contextual, findable, not dependent on any single person’s availability or generosity.

This is not about making your community “AI-ready,” a phrase I’d actually be gloriously happy to retire. Communities serve humans. Full stop. But humans now move through information differently. They come in with partial answers. They paste things into tools and come back with follow-up questions. They expect to find what they’re looking for without having to know who to ask or which part of your site to go to. That’s not a new expectation, actually. It’s just newly consequential when it goes unmet.

So what does this mean practically? It means documentation that stands alone, with enough surrounding context that a reader (human or otherwise) doesn’t need backstory to make use of it. It means tagging and titling that reflects how people ask questions, not how the internal team thinks about categories (I’m glaring at you personas). It means periodically asking: if Jillian left tomorrow, what would actually be findable? What would disappear with her? (Not much actually; Jillian knows to document her brain on the regular but Jillian is weird.)

It also means letting go of the fantasy that relational knowledge transfer scales. It doesn’t. It never did. Jillian has been subsidizing your knowledge architecture for years, and she deserves a vacation. Italy or France will do.

The communities worth building, right now, are the ones that work when nobody knows anybody. They’re also the ones that become richer when people do connect, when Jillian’s expertise gets documented and built upon rather than just tapped and forgotten.

And y’all… that’s not an AI story. That’s just a community worth having.

Go give Jillian some time off.