An automotive knowledge base becomes less useful when it remembers the conclusion but loses the source, date and uncertainty that made the conclusion meaningful.
Our April discussions about extracting intelligence from articles, posts and transcripts addressed that risk directly. The purpose was to build reusable understanding, not an expanding collection of attractive summaries.
Vertex Automotive Solutions would retain the original signal, its author and publication context, then distinguish the Vertex interpretation from the source's claims. A supplier announcement, independent study and consultant opinion should not acquire identical evidential weight because they enter the same system.
Time matters too. A proposal may later become a service, a market estimate may be revised and an operational recommendation may depend on a policy that has changed. Retrieval needs to preserve enough context for the user to judge whether the evidence still applies.
Omnia's knowledge direction is to support a better decision with traceable material. Jano should be able to explain the relevant source and its limits, or say when the record cannot support the answer. Adding more documents does not establish that this standard has been met.
Test the knowledge base with a question containing conflicting evidence. If it cannot distinguish the sources and explain the uncertainty, a larger collection is unlikely to resolve the underlying weakness.