02 — What Modern Knowledge Management means

The difference is one layer — and where you put it.

Most organisations connect AI straight to their documents and hope for the best. The model reads what it's given and produces the most plausible answer. It works, right up until the moment the answer matters.

Modern Knowledge Management puts a layer in between: an explicit model of the things you deal in, and a graph of how they relate — each statement carrying its source. The AI then works on that, instead of guessing.

The common approach
Documents & systems catalogue · reports · files · spreadsheets · email AI model READS · PREDICTS · WRITES A fluent, confident answer NO TRAIL BACK · NOT REPRODUCIBLE
Modern Knowledge Management
Documents & systems the same material you already hold THE KNOWLEDGE LAYER · YOU OWN IT Ontology + knowledge graph what is known — and how it came to be known AI model READS THE GRAPH · ASSEMBLES ANSWER + ITS EVIDENCE · REPRODUCIBLE
The whole argument in one picture. Same documents, same model. The difference is whether the institution's knowledge lives in a layer it governs, or only inside the model's guess.

So what is actually in there?

Less exotic than it sounds. It's the diagram anyone would draw on a whiteboard to explain how things in your institution relate — written down in a form a machine can follow.

Anatomy — three parts, and you already think this way
Bowl OBJECT S. Wattana PERSON 1911 trip EVENT WAS COLLECTED BY TOOK PART IN PROPERTIES ON THE THING material · dimensions · location accession number · condition PROPERTIES ON THE RELATIONSHIP source: donor correspondence status: corroborated · asserted by: J.P. AND SO ON, WITHOUT LIMIT new kinds of thing attach to what exists — nothing gets rebuilt THINGS BECOME NODES · HOW THEY RELATE BECOMES A NAMED LINK · EVIDENCE RIDES ON THE LINK ITSELF THE SECOND BOX IS THE UNUSUAL PART — MOST SYSTEMS RECORD THAT TWO THINGS ARE RELATED, BUT NOT HOW ANYONE KNOWS
The relationship carries the evidence. That single decision is what makes the whole thing auditable later — and it is the part almost every data model leaves out.
And it remembers three different things, not one
01 · WHAT IS BEING ASKED NOW The live enquiry the conversation in progress — what a visitor or researcher has already been told today 02 · WHAT THE INSTITUTION KNOWS The settled record objects, people, places, events — organised by your ontology, each statement with its source 03 · WHY IT CONCLUDED THAT The reasoning record which evidence was weighed, what was rejected and why, who decided, and when PRECEDENT MADE QUERYABLE — THE NEXT SIMILAR CASE STARTS FROM WHAT WAS DECIDED LAST TIME MOST SYSTEMS STORE ONLY THE MIDDLE ONE — AND KEEP THE CONCLUSION WITHOUT THE ARGUMENT THE THIRD IS THE ONE THAT USUALLY RETIRES WITH THE PERSON WHO HELD IT
The third one is the quiet loss. When an attribution is revisited in twenty years, the useful thing is not the conclusion — it's knowing what evidence was considered, what was rejected, and why.
Old knowledge management

Store it so a person can find it

Repositories, folder structures, search. Optimised for retrieval by a human who already knows roughly what they're looking for.

What changed

A machine now answers first

People ask a model before they open a system. Whatever the model can't verify, it invents — fluently, and in your institution's name.

Modern knowledge management

Structure it so a machine can be held to it

Explicit meaning, explicit relationships, explicit evidence. Built so the answer can be traced, repeated and defended.

Start the conversation

If this is close to what you have been thinking, let's find a time.

90 minutes. Bring your curators or librarians, and one question your institution couldn't answer.

Received. A confirmation is on its way to your inbox — we'll reply within two business days to find a time.

Or write directly — hello@neogens.co

Previous
The problem
Next
Why it works