IskayPet builds corporate knowledge assistant on Amazon Bedrock Managed Knowledge Bases

By using Bedrock’s managed RAG stack, IskayPet avoided running its own vector database, embedding pipeline or ingestion jobs. The native SharePoint connector imported documents with advanced indexing, including text from images, and document‑level permissions were enforced at query time. The system retrieves up to 25 passages, injects up to 15 into a prompt for Claude Haiku 4.5, and returns a cited answer.

In the first two months of production, answer‑finding time dropped from 3–5 minutes to about 7 seconds, and employees spent roughly 95 percent less time searching for procedures. The assistant now handles roughly 100,000 queries per month across all business areas and countries.