Your company's best knowledge is locked inside someone's head. Here's how to free it.

There is a person at your company who knows exactly why the payments integration was built the way it was. There is someone who has seen this customer complaint pattern before, twice, and knows what to do. There is an engineer who spent three months learning a framework that three other engineers are about to start learning from scratch.

The knowledge exists. The problem is discoverability.

Most organisations approach this problem with documentation. Write it down, they say. Put it in Confluence. Add it to the wiki. Build a knowledge base.

And so the wiki grows - thousands of pages, many outdated, almost none findable in the moment someone actually needs them - until searching it becomes a skill in itself, and most people give up and ask in Slack instead.

The human search engine problem.

When someone needs to know something at work, they do one of two things. They search a document repository and find nothing useful. Or they ask a person.

Asking a person is almost always better. People give context. They know what you're really asking even when you haven't quite articulated it. They tell you the official answer and the thing you actually need to know. They point you to the right next person if they can't help.

The problem with asking a person is finding the right person. This is harder than it sounds.

In a 20-person company, you can learn who knows what through proximity and pattern - you overhear conversations, you're in the same room, you notice who people go to when they need a decision made. In a 200-person company, the knowledge is distributed across people you've never met, in timezones you don't think about, with skills and experiences that aren't visible anywhere.

The question "who should I ask about this?" becomes genuinely hard. And most organisations have no answer.

The taxonomy problem.

Even when companies try to document expertise, they run into the taxonomy problem: how do you organise knowledge in a way that matches how people search for it?

Skills databases fail because they're built around job titles and standard competencies - "Project Management: Advanced," "Excel: Intermediate" - rather than the specific, contextual knowledge that's actually useful. Nobody searches for "Excel: Intermediate." They search for "who knows how to model variable commissions in a spreadsheet."

The answer isn't a better taxonomy. It's a better search.

Natural language as the interface.

The shift that makes people-finding genuinely useful is the move from browsing and filtering to asking.

Instead of a directory with dropdowns - department, location, skill level - an interface where you type what you actually want to know: "Who has experience building integrations with Salesforce?" "Who's done enterprise sales in the German market?" "Is there anyone who's worked with this library before?"

This works when two things are true. First, when the underlying data is structured - not a free-text bio, but specific skills, projects and experiences captured as queryable fields. Second, when a language model can map what you're asking to what's actually in the data, handling the variation in how people phrase things and the gaps in how they describe themselves.

The result is that institutional knowledge stops being something you have to know to ask about. You can ask what you're looking for, and find the person who can help.

The knowledge that walks out the door.

There's a harder version of this problem, which is what happens when people leave.

When someone who's been at a company for four years hands in their notice, they take with them an enormous amount of context, judgment and institutional memory that was never captured anywhere. The decisions they made and why. The patterns they'd learned to recognise. The relationships they'd built with customers and counterparts.

Some of this can be documented during an offboarding process. Most of it cannot - not because it's secret, but because it's not the kind of knowledge that transfers through documentation. It transfers through conversation, mentorship and ongoing access to the person.

The best organisations approach this proactively: identifying the people who hold critical knowledge, making sure that knowledge is distributed rather than concentrated, and creating the conditions - a good directory, a culture of coffee chats and mentorship - where knowledge is continuously flowing rather than sitting in one person's head waiting for the wrong moment to become critical.