Your organisation has a knowledge problem. You just don't call it that.

The real cost of institutional knowledge trapped in documents, inboxes, and people's heads, and why most organisations don't recognise it until someone leaves.

Your organisation has a knowledge problem. You just dont call it that.

Every organisation I work with has the same problem. They just describe it differently.

The law firm says they need to find precedents faster. The hospital says clinicians are spending too long reviewing patient history. The NGO says their field teams cant access the right guidance quickly enough. The corporate ops team says reporting takes too long because nobody can find the source data.

Different industries. Different language. Same problem.

The knowledge your organisation depends on, the experience, the decisions, the precedents, the process documentation, exists somewhere. But "somewhere" is doing a lot of heavy lifting. It might mean an inbox. A shared drive nobody has organised in three years. A folder structure that made sense to the person who built it and nobody else. Or it might mean asking someone who has been there long enough to remember.

The cost isnt storage. Its the hours.

Most organisations think about knowledge as a storage problem. Where do we put things? How do we organise them? What filing system do we use?

That is the wrong frame entirely.

The cost of a knowledge problem is not storage. Its retrieval. Its the two hours a lawyer spends searching through case archives to find a precedent they know exists. Its the nurse reading through clinical notes that a previous shift already documented. Its the analyst rebuilding a report that someone else built six months ago because they didnt know it already existed.

This is invisible cost. It doesnt show up as a line item. It shows up as people being slower than they should be, and nobody being able to explain exactly why.

Why it stays invisible

There are a few reasons organisations dont recognise this as a systems problem.

First, the people doing the searching dont flag it. They assume it is just part of the job. They have always had to dig through folders. That is what searching for things feels like.

Second, the cost is diffuse. Its not one person losing one hour. Its thirty people losing twenty minutes each, every day, across every team. That is invisible at the individual level and enormous at the organisational level.

Third, organisations confuse having information with being able to access it. Your documents are all there. Your emails are archived. Your case notes are in the system. The assumption is that if it was recorded, it is accessible. It is not. Having data and being able to use it quickly are two completely different things.

What this looks like by sector

Every sector has its own version of this problem. Here are the ones I see most often.

Law firms have decades of precedent locked in document management systems that support keyword search at best. Lawyers know roughly what they are looking for, but the query that finds it requires knowing the exact words used in a document written by a partner who left in 2018.

Hospitals and clinical teams have patient records, clinical guidelines, treatment protocols, and research buried across multiple systems. Clinicians spend a material percentage of their time retrieving and synthesising information that the system theoretically contains.

NGOs and international organisations have field reports, project evaluations, and institutional knowledge spread across country offices, staff turnover, and disconnected file stores. Every new programme officer partially reinvents what the previous one already figured out.

Corporate operations teams have process documentation written when a process was built, updated inconsistently, and frequently ignored because it takes longer to find the document than to ask a colleague.

The "ask someone" fallacy

The most dangerous workaround is the one that feels the most natural: just ask someone who knows.

This works. Until it doesnt.

It works when the person who knows is available, is still employed here, and has time to explain. It stops working when they leave, are on leave, or are in a different timezone. It also stops working at scale, when the volume of questions exceeds the capacity of the people who hold the answers.

Organisations that run on "ask someone who knows" are not just slow. They are fragile. One departure and years of institutional knowledge walk out the door.

I have seen this play out more than once. A senior person leaves. Suddenly nobody knows how a critical process actually works, because it was never written down. Or it was written down, but not in a way that is findable. The result is months of relearning what already existed.

What the fix looks like

The fix is not a new folder structure. It is not a better naming convention. It is not mandating that people update the wiki they never update.

The fix is making knowledge queryable.

Retrieval-Augmented Generation (RAG) is the technology that makes this practical. You take the documents, emails, reports, case notes, and policy files your organisation already has, and you build a system that lets people ask questions in plain language and get accurate, referenced answers back.

The result is not magic. It is: your team stops searching and starts asking. And the system finds what they need in seconds instead of hours.

I built a version of this for a bid team that was spending days manually searching through past winning proposals every time they had to write a new one. The documents all existed. They just werent accessible in any practical sense. After building a RAG-powered tool over their document library, the research that used to take two days took twenty minutes.

The documents didnt change. The accessibility did.

The honest bottom line

If your team regularly says things like "I know we have something on this somewhere" or "let me ask so-and-so, they will know," then you have a knowledge problem.

The good news is that solving it does not require replacing your systems or changing how people work. It requires making what you already have actually findable.

That is a tractable engineering problem. And it tends to have one of the highest returns of any AI project I work on. The cost it is solving is enormous, it is just invisible.