Amazon Bedrock Managed Knowledge Base launches fully managed enterprise search for agents

Amazon Bedrock Managed Knowledge Base is now generally available, offering a fully managed solution for grounding generative AI agents on enterprise data. The service abstracts the complexity of building custom data ingestion pipelines, managing vector or graph databases, and orchestrating retrieval. Developers can connect enterprise data sources—including Amazon S3, Microsoft SharePoint, Atlassian Confluence, Google Drive, and Microsoft OneDrive—or crawl the web, with sensible defaults enabling first retrieval in minutes. The service automatically handles vector store scaling, multimodal parsing (PDFs, PPTX, DOCX, scanned docs, audio, video), and ongoing syncs that only process modified documents.

Production readiness is a key pillar: Managed Knowledge Base implements real-time ACL checks by querying the authoritative source at retrieval time, rather than relying on stale cached permissions. This provides an additional security layer beyond pre-retrieval filtering. Users retain control over embedding models, rerankers, and chunking strategies when needed. The solution already powers internal AI copilots at companies like Syngenta Group and MRH Trowe, enabling secure, grounded answers across thousands of documents in multiple languages without building custom pipelines.