Veylund internal system
AI knowledge & decision system
Durable Context
Problem
Finding a document does not tell a new colleague or AI session whether it is current, why a decision changed or who owns the next action.
Approach
Link sources to decisions and owners. Separate current guidance from historical context, preserving the reasons for corrections.
What exists today
An internal knowledge system in development. The diagrams illustrate its approach, not the private implementation.
Project walkthrough
Connect an answer to its evidence
Start with the source material, the decision it supports and the responsibility attached to it. The goal is a reviewable chain, not an unsupported summary.
Separate retrieval from continuity
A conversation holds recent context; retrieval can bring relevant material into a task. Continuity adds a further concern: how decisions, corrections and ownership remain understandable over time.
Preserve what changed and why
When a decision is corrected, the current position should be clear without erasing the history that explains it. That distinction supports handovers and subsequent review.
Design rationale
A reviewer can check both the source of an answer and the history behind the current decision.
Related service