Amazon Bedrock Knowledge Base Adds Salesforce and Zendesk Connectors

AWS deepened its enterprise agentic and RAG stack with native Salesforce and Zendesk connectors for Amazon Bedrock Managed Knowledge Base. The connectors let customers sync Salesforce and Zendesk knowledge articles and community posts directly into Bedrock's managed retrieval-augmented generation service, removing the need to build and maintain custom ingestion pipelines — a common friction point in deploying enterprise support and knowledge assistants.

The release is part of a steady drumbeat of AWS enterprise-AI plumbing this week: guidance on pairing open-weight models with the open-source OpenCode terminal agent on Bedrock, an agentic conversational video-intelligence reference architecture using the Strands Agents SDK to orchestrate Bedrock, Rekognition, and Transcribe, and Dutch retailer HEMA's internal assistant HAL built on Bedrock AgentCore with Model Context Protocol. AWS's VP of Agentic AI, Swami Sivasubramanian, also announced the open-source Strands harness, claiming frontier agent performance at 28% lower token cost.

The strategic thread is unmistakable: AWS is building the connective tissue for enterprises to wire their existing systems of record (CRM, support, video) into governed, model-agnostic agentic applications. Rather than competing on frontier models, AWS competes on integration breadth, governance, and cost efficiency — the Salesforce and Zendesk connectors are a small but concrete example of reducing the integration tax that slows enterprise AI adoption. For teams building support-desk or internal-knowledge assistants, native connectors mean faster time-to-value and less brittle custom code; the trade-off is deeper lock-in to the Bedrock ecosystem. Watch for more first-party connectors (ServiceNow, Workday, SharePoint) as AWS races Microsoft's Foundry and Google's Vertex on enterprise data reach.