AWS launches a knowledge graph that learns from AI agent activity

Amazon Web Services has announced a new set of tools designed to give AI agents better access to company data. The centerpiece is AWS Context, a knowledge graph service that automatically maps relationships across an organization’s data and improves over time based on how agents actually use it. Sean Michael Kerner reports for VentureBeat that the announcement came alongside two additional services: Amazon S3 Annotations and skill assets in AWS Glue Data Catalog.

Swami Sivasubramanian, AWS vice president of Agentic AI, described the core promise at the AWS Summit NYC: “This service automatically builds a knowledge graph from all your existing data. This service infers relationships across your data sets, business rules, and domain knowledge, and makes all of it available to your agents and your organization at runtime.”

The key distinction AWS is making is that the graph does not require manual curation. It learns which data sources produce correct results and which parts get used most. Human data stewards can still review and approve inferred relationships through the AWS Management Console, but the system is designed to reduce the ongoing maintenance burden.

Three layers, one system

• AWS Context builds and maintains the knowledge graph, combining semantic search with graph-level reasoning across structured and unstructured data sources.
• Amazon S3 Annotations lets users attach business context directly to individual files in Amazon S3 storage.
• AWS Glue Data Catalog skill assets link runbooks, query patterns, and usage rules to data assets across a company’s environment.

AWS also unveiled a new security agent called AWS Continuum, which starts in a supervised mode and can be granted increasing autonomy to fix code vulnerabilities on its own. Todd Bishop reports for GeekWire that the broader theme across all announcements is maximizing what AI can do while keeping humans in control of final decisions.

AWS enters a crowded field. Snowflake, Microsoft, Redis, and Pinecone have each introduced their own approaches to organizing enterprise data for AI agents. Analyst Holger Mueller of Constellation Research told VentureBeat: “Every agentic platform vendor needs a context capability.” He also flagged a shared concern: “The concern, as with all context offerings, is going to be performance, especially for transactional data.”

Sources

• AWS Context knowledge graph learns from agents – VentureBeat
• Amazon unveils new AI agents, trying to thread the needle between autonomy and human control – GeekWire