AWS ML BlogThursday · October 1, 2026FREEQuery claims in natural language with Amazon Bedrock Knowledge Bases
AWS's Machine Learning Blog published a technical how-to on building a conversational claims assistant using Amazon Bedrock Knowledge Bases. According to the excerpt, the assistant answers natural-language questions and returns citations with its answers. The post walks through ingesting claim documents from Amazon S3, then querying them with the AgenticRetrieveStream API. It also covers multi-turn follow-ups, metadata filters, and contextual grounding guardrails. The excerpt frames the piece as a technical how-to rather than a product announcement, and it does not state pricing, benchmark results, model versions, or availability details. The described components are limited to Amazon Bedrock Knowledge Bases, Amazon S3 as the document source, the AgenticRetrieveStream API for querying, and the supporting features of multi-turn follow-ups, metadata filters, and contextual grounding guardrails. No performance figures, latency numbers, or accuracy claims appear in the provided text.
Developers building retrieval-based assistants can follow the post's pattern for querying claim documents with citations using Bedrock Knowledge Bases and the AgenticRetrieveStream API.