Stardog and AWS Bedrock AgentCore build semantic layer for agentic AI analytics

Stardog and Amazon Bedrock AgentCore have collaborated to build a semantic layer for agentic AI on AWS, addressing the long-standing enterprise goal of shrinking the time between a business question and a trustworthy answer. Traditional analytics progressed from scheduled reports to dashboards to self-service BI, but all required a prepared dataset. Generative AI agents now promise to autonomously reason over live data by planning, writing queries, evaluating results, and iterating—yet they struggle when enterprise data is fragmented across systems like Amazon Aurora (operational) and Amazon Redshift (analytics), where definitions of "customer" or "revenue" differ. Stardog's Semantic AI Application captures business context once and makes it reusable, enabling agents to compose answers that are consistent and auditable without Extract, Transform, Load (ETL) pipelines.

On the technical side, the semantic layer sits behind Strands Agents operating on Amazon Bedrock AgentCore, which bundles inbound authentication, hosting, and tool credentials into one managed service. Foundation models available on Bedrock—like those providing multi-step reasoning—are already capable of producing valid SQL, but the missing piece is the business context: shared definitions of entities, rules, and metrics. The Stardog layer fills that gap, allowing agents to query across Amazon Aurora and Amazon Redshift with the same proficiency as a senior analyst. The same deployment also works behind Amazon EKS, Amazon ECS, and AWS Lambda. Complementing Retrieval-Augmented Generation (RAG) for unstructured text, this semantic layer handles analytical questions that require joining live records, applying consistent business logic, and honoring row- or column-level access policies—a critical step toward trustworthy agentic analytics.