AWS unveils Amazon Bedrock Managed Knowledge Base to simplify enterprise RAG pipelines

Amazon Bedrock Managed Knowledge Base is a new capability in Amazon Bedrock designed to let developers build enterprise-grade generative AI applications using proprietary data with minimal infrastructure management. The service abstracts retrieval, embeddings, re-ranking, and foundation model selection into a single managed primitive, allowing teams to focus on business outcomes rather than assembling and maintaining RAG components.

To address common enterprise challenges, the service provides six native data connectors at launch: Amazon S3, SharePoint, Confluence, Web Crawler, Google Drive, and OneDrive. These connectors pull enterprise data and associated permissions directly into the knowledge base, eliminating the need for custom connectors and reducing setup complexity.

Smart Parsing automates multi-format data preparation by selecting optimal parsing strategies for each data type and connector. It preserves document structure, handles multimodal content such as images and tables, and applies foundation-model-powered chunking to improve retrieval accuracy without manual experimentation.

Agentic Retriever enables complex, multi-turn queries by decomposing user intent into step-by-step plans and performing multi-hop retrieval across one or more knowledge bases. This supports scenarios where a single retrieval step would otherwise miss context, such as combining budget information with policy documents to answer compound questions.

The service integrates with Amazon Bedrock AgentCore Gateway, allowing developers to attach a managed knowledge base to an agent with a few lines of code. AWS automatically generates role-based permissions and exposes observability and evaluation metrics through the AgentCore Observability dashboard, providing visibility into performance and usage.