The library is fragmented
Documents, systems, email, meeting notes and individual experience contain different parts of the story.
AI cannot organise the truth if the library is still a mess.
Most organisations are asking powerful AI to work across years of duplicated files, conflicting versions, obsolete decisions, missing context and knowledge that exists only in people’s heads.
The uncomfortable reality
Better AI does not make poor knowledge reliable.
Retrieval can find passages. A larger context window can hold more material. Neither decides which source is authoritative, whether a position has been superseded, why an expert reached a conclusion or whether that conclusion applies here.
AI receives more noise
Large volumes of weakly controlled material consume context and can obscure the small amount that matters.
The answer sounds complete
Fluent output can hide uncertainty, conflicting evidence, missing ownership and obsolete information.
People still cannot trust it
The user cannot easily see what is current, who approved it, what supports it or where it applies.
The token problem is a knowledge problem
Sending more of the library is not the same as sending better knowledge.
Repeatedly supplying large quantities of irrelevant or duplicated source material can increase token usage and processing cost. More importantly, it increases the burden of finding the authoritative interpretation inside the noise.
Library first
- Retrieve broad collections of documents
- Mix current and superseded material
- Reconstruct context for every request
- Depend on the model to infer authority
- Repeat the same interpretation work
Knowledge first
- Retain sources for evidence and provenance
- Establish the current knowledge position
- Attach ownership, status and context
- Expose uncertainty and review requirements
- Provide a smaller, relevant layer for use
The library and the doctor
The doctor does not prescribe the medical library.
▤
The library contains information
It includes evidence, history, disagreement, old practice and material of uneven authority.
✚
The doctor applies expertise
The professional interprets evidence in the context of the patient, applies judgement and accepts responsibility.
◉
Interrogate captures the reasoning
Knowledge is separated from raw material while remaining connected to its provenance and challenge.
✦
AI receives what matters
The knowledge layer can provide relevant, governed context without pretending that automation replaces authority.
A more useful AI foundation
Build knowledge once. Apply it many times.
Current positions are visible
People can distinguish established knowledge from source material, draft thinking and unresolved questions.
Specialist judgement scales
Expert interpretation becomes available to authorised people without requiring the expert to answer every question again.
AI answers are inspectable
Relevant provenance, ownership and context can remain available for review.
Knowledge keeps evolving
Changes in evidence or circumstances can trigger focused challenge and review rather than another uncontrolled document dump.
Do not build AI around the mess. Build the knowledge layer that lets AI become useful.
Interrogate by Innovate Tax
Give people and AI the knowledge that matters now.
Start with the decisions where fragmented information and scarce expertise create the greatest risk or delay.