AWS Joins the Context Layer Competition with an AI-Driven Graph Learning from Agents, Not Manual Input

The creation of a contextual layer that connects enterprise data systems with AI agents is a specialized task, lacking a standardized service for either automation or ongoing maintenance of data graphs. Amazon aims to change this landscape.

On Wednesday, Amazon announced its entry into the market with the introduction of three new products designed as a context intelligence stack tailored for AI agents. The highlight of this offering is AWS Context, a new service that develops a knowledge graph that enhances itself through agent interactions over time. Additionally, Amazon announced the general launch of Amazon S3 Annotations and a preview of skill assets within the AWS Glue Data Catalog.

The context layer has become an increasingly competitive segment, with various vendors offering solutions. AWS’s approach diverges by proposing that the knowledge graph should learn autonomously from agent usage, eliminating the need for manual updates.

“Your agents can become smarter without the need for extensive rebuilding,” remarked Swami Sivasubramanian, VP of Agentic AI at AWS, during his keynote at the AWS Summit in NYC.

He added, “This service constructs a knowledge graph derived from your existing data, drawing inferences about relationships between datasets, business rules, and domain knowledge, making it readily accessible for your agents and organization in real-time.”

AWS Context: A Self-Improving Knowledge Graph

AWS asserts that this challenge has frequently surfaced in customer implementations.

AWS Context automatically identifies and maps relationships in existing data, including table structures, column meanings, authoritative sources, and interrelations. It merges semantic search capabilities with graph reasoning, allowing for inferred relationships to be made available to agents as they operate.

“The knowledge graph evolves continually as it learns which sources yield accurate results and which aspects are utilized,” Sivasubramanian stated.

Data stewards can oversee the graph using the AWS Management Console, where they can assess inferred relationships, upgrade them to production status, and add business definitions and usage guidelines. Every query is aligned with the user’s IAM and Lake Formation permissions, ensuring traceability for data access through established enterprise controls.

All metadata is stored in Apache Iceberg format within Amazon S3 Tables and is queryable through platforms like Athena, Redshift, Spark, or any engine compatible with Iceberg, eliminating the need for proprietary APIs. The service supports connections to external catalogs, allowing context from non-AWS systems to be integrated into the knowledge graph. Agents can utilize agentic search APIs and MCP tools via Bedrock AgentCore, EKS, or any compatible framework.

More Than Just a Single Service

The realm of context is intricate, and AWS is building an array of services to aid enterprises in establishing context throughout their data stack.

Amazon S3 Annotations: This service allows users to append rich business context directly to individual S3 storage objects.

AWS Glue Data Catalog Skill Assets: This feature links domain knowledge to data assets through runbooks, query patterns, and usage rules within the catalog layer.

AWS Context synthesizes these elements into the knowledge graph that agents interact with during run time, combining semantic search with graph reasoning across both structured and unstructured data sources. Each service layer complements the next.

Entering a Competitive Field

This month, Snowflake rolled out its own context approach with Horizon Context and Cortex Sense services. Microsoft offers a context feature through its Fabric IQ platform, which includes a semantic ontology for data management. Redis has introduced a context platform optimized for data retrieval, while Pinecone, a vector database provider, presents its Nexus context solution for compiling enterprise data into specific task artifacts before agents query them.

AWS’s structural advantage is clear: for enterprises utilizing S3, Glue, and Lake Formation, AWS Context provides an extension to an existing identity framework with minimal data movement. This proposition emphasizes low integration challenges rather than simply cost efficiency.

“Context empowers agents, and as the global focus shifts towards developing agents, every platform vendor in this space will need context capabilities,” noted Holger Mueller, VP and Principal Analyst at Constellation Research.

Mueller acknowledged AWS’s involvement but raised concerns, stating, “The primary challenge — like any context solution — will be performance, particularly concerning transactional data, which remains to be seen.”