AWS Launches Amazon Bedrock Managed Knowledge Bases: A New Tool for Enterprise Knowledge Management

Summary:AWS has launched Amazon Bedrock Managed Knowledge Bases, a RAG-based enterprise knowledge management solution that eliminates the need for businesses to manage vector databases or embedding pipelines. The service integrates seamlessly with existing data sources like SharePoint, supports multilingual and multi-domain knowledge retrieval, and offers document-level access control and hybrid search capabilities. IskayPet's use case demonstrates its effectiveness in retail, veterinary, onboarding, an

---

Key Breakthroughs

AWS has launched Amazon Bedrock Managed Knowledge Bases, a RAG-based enterprise knowledge management solution that eliminates the need for businesses to manage vector databases or embedding pipelines. The key features of this service include:

1. Managed RAG Service: Amazon Bedrock handles data ingestion, multimodal parsing, vector storage, embedding generation, and retrieval. Businesses only pay for what they store and retrieve, without managing the underlying infrastructure.
2. Native SharePoint Connector: Seamlessly integrates with existing SharePoint environments without data migration and supports processing image content embedded in documents.
3. Hybrid Search: Combines semantic and keyword search (BM25) to ensure the accuracy and comprehensiveness of retrieval results.
4. Document-Level Access Control: Enforces SharePoint permissions to ensure users only access content they are authorized to view.
5. Model Flexibility: Businesses can choose different cost models and switch between them without restructuring the system architecture.

Technical Architecture and Implementation

Amazon Bedrock Managed Knowledge Bases adopts a modular design, allowing businesses to quickly build frontend applications using tools like Kiro and leverage models like Claude Haiku 4.5 for natural language processing tasks. The main technical components include:

• Frontend Building: Use Kiro to build employee-facing application interfaces and generate and iterate UI code with Claude Haiku 4.5.
• Knowledge Base Configuration: Each knowledge base can be configured for specific SharePoint sites or document libraries to ensure only target content is indexed.
• Retrieval and Generation: Employs a hybrid search strategy combining semantic and keyword retrieval and uses Amazon Bedrock Guardrails for content filtering and responsible AI policy application.

Engineering Trade-offs and Optimizations

• Cost Control: The main costs come from index storage, retrieval API calls, and generation model token usage. The team is evaluating optimizations such as prompt caching and batch inference to control costs.
• Performance Improvement: By adjusting the number of retrieval results and generation prompt settings, the system maintains answer quality while controlling computational overhead.
• Multilingual Support: Supports retrieval and generation in four languages without maintaining separate systems.

Real-World Application Results

The IskayPet case demonstrates that Amazon Bedrock Managed Knowledge Bases significantly improves enterprise knowledge retrieval efficiency:

• Query Speed: Reduced from 3-5 minutes to about 7 seconds.
• Productivity Boost: Search time reduced by 95%.
• Scalability: Expected to handle about 100,000 queries per month.

Developer Recommendations

For businesses facing similar knowledge management challenges, it is recommended to start with the AWS Free Tier and build the first knowledge base using the Amazon Bedrock console. Additionally, the following resources can be referenced:

• Amazon Bedrock Managed Knowledge Base
• Evaluate the performance of RAG sources using Amazon Bedrock evaluations

Conclusion

Amazon Bedrock Managed Knowledge Bases provides an efficient, scalable, and cost-effective knowledge management solution for enterprises. Through its managed RAG service, native SharePoint connector, and hybrid search capabilities, businesses can quickly deploy knowledge management systems to enhance employee productivity and decision-making capabilities.

---

Source: AWS Machine Learning Blog (2026-10-22)

— END —

Tags: #AWS Follow #Amazon Bedrock Follow #RAG Follow #Knowledge Management Follow #Multilingual Support Follow