Lebendiges Wissensmanagement durch KI-Agenten im KMU-Vertrieb -konzeption und Evaluation eines selbst-aktualisierenden Wissenssystems

A prototype called “Living Brain” was developed to enable continuous, self‑updating knowledge management for small‑ and medium‑sized insurance distributors by using AI agents that automatically ingest documents, rank results with quality‑aware reranking, and close a feedback loop that flags outdated information and detects knowledge gaps. Controlled experiments demonstrated that the adaptive reranking reduced weak retrieval hits by more than 50 % and that the system reliably identified missing knowledge with high accuracy, while its performance was shown to depend on the chosen embedding model and domain coverage.