AWS latest to introduce context layer for agentic AI
With context key to agents performing in production, the tech giant's latest new features aim to feed cutting-edge tools the situational awareness they require.
AWS on Wednesday joined the cavalcade of data platform providers to unveil capabilities designed to connect agentic AI tools with the context they require to properly perform in production.
Context such as an enterprise's proprietary data and business logic, including organizational rules and the results of past analyses, give agents the situational awareness to act autonomously, aiding individual workers and sometimes executing entire business processes. Without relevant context, agents will make inferences based on the information they do possess, which can lead to inaccurate outputs.
AWS Context, unveiled during AWS New York Summit, a user conference hosted by the tech giant, is a service that automatically maps related data and business logic into a knowledge graph that agentic AI applications can search and discover to inform outputs.
Given that AWS Context addresses a core requirement for agentic AI, the new service will be valuable for AWS users once it's made generally available, according to Jake Dolezal, lead data engineer at McKnight Consulting Group, who noted that AWS Context will enable organizations to improve on pre-existing data retrieval pipelines that have often proven insufficient for agentic AI.
"It's a meaningful shift from each team building its own [retrieval-augmented generation] pipeline to one governed context layer the whole organization draws from," he told TechTarget.
Beyond AWS Context, additional features aimed at connecting agentic AI tools with context include business context and semantic search for AWS Glue Data Catalog and Amazon S3 annotations in AWS' Simple Storage Service for object storage.
Context is key
Agentic AI development has been a significant initiative for many organizations since agents emerged as the cutting-edge of enterprise AI in 2024. Simultaneously, with the data and business logic agents require to understand the unique characteristics of an organization stored in their platforms, many data management and analytics vendors have created development environments aimed at making it easy for customers to build agents informed by proprietary information.
It's a meaningful shift from each team building its own [retrieval-augmented generation] pipeline to one governed context layer the whole organization draws from.
Jake DolezalLead data engineer, McKnight Consulting Group
However, despite the overwhelming emphasis on AI development in recent years, studies such as a 2026 survey by Deloitte show that most AI initiatives fail to make it into production. Factors such as technology and talent deficits and lack of organizational buy-in are factors. But perhaps the biggest barriers enterprises face are trying to organize vast data estates to make data AI-ready, and finding and operationalizing relevant data to feed agents.
In response, in 2026 a slew of data and analytics vendors have introduced tools aimed at connecting agents with context. A partial list includes leading hyperscale, data platform, database and analytics vendors including Google Cloud, Microsoft, Databricks, Snowflake, MongoDB and Tableau.
Now, AWS is among them.
"You want to trust the decisions made by your AI agents, but that can't happen until agents have context," AWS vice president of technology for data analytics Mai-Lan Tomsen Bukovec wrote in a blog post. "Imagine what becomes possible when we give agents a safe way to access the context they need to deliver trusted decisions."
AWS Context is an extension of the knowledge graph technology that powers Amazon Quick, an agentic AI assistant for business employees. The knowledge graph catalogs datasets, dashboards and metadata. In addition, it learns from usage patterns to improve the relevancy of responses. But rather than serve individuals as in the case of Amazon Quick, AWS Context's knowledge graph informs agentic AI applications across entire organizations.
In addition, AWS Context includes built-in governance to restrict the data agentic AI tools can access and show what agents accessed and under whose authority it did so. AWS Context also makes all queries identity-aware so that an agent can only see and operationalize information it is authorized to access.
Like Dolezal, Donald Farmer, founder and principal of TreeHive Strategy, noted that the new service addresses a pressing need.
"The core claim that AI agents fail because they lack the context to reason reliably over enterprise data is well recognized," he told TechTarget. "The best idea in the announcement is the identity-aware query design. Tying agent access to identity and access management permissions so that an agent inherits only the correct authorizations is a solid architectural response to a real problem in governance."
Enabling the knowledge graph to learn from usage patterns is a valuable feature, Farmer continued, while adding a warning.
"I like the knowledge graph learning from usage, but with some caveats," he said. "AWS says the graph observes which join paths agents rely on and propagates correct patterns across the organization without human re-curation. Well, that assumes the agents are getting things right."
Agents, especially those without enough relevant context, are imperfect, Farmer noted. Just as they can follow correct patterns, they can follow incorrect join paths and draw on incorrect data sources.
"A graph that learns from those interactions would replicate and distribute the errors without human intervention and possibly without even being noticed," he said. "I'd like to know what feedback signals help us correct for this."
Supplementing context
While AWS Context provides a knowledge graph to inform agentic AI, business context and semantic search for AWS Glue Data Catalog augments the new service. The feature, which is in preview, enables users to enrich tables, views and columns in their catalog with business descriptions, glossary terms and custom metadata so they provide added context to agents.
Meanwhile, Amazon S3 annotations, which is now generally available, enables users to attach business context directly to S3 objects that can be stored in Apache Iceberg tables within AWS' S3 object storage service.
Both business context and semantic search for AWS Glue Data Catalog and Amazon S3 annotations represent strategic infrastructure expansion, according to Dolezal.
"The Glue Catalog enhancements and S3 annotations are incremental improvements that fill real gaps," he said.
In particular, S3 annotations, which allows users to attach up to 1 GB of context to S3 objects that automatically flow into Iceberg tables, is significant, Dolezal continued.
"It's not flashy, but it removes genuine friction that every data platform team currently owns manually," he said.
Looking ahead, as AWS continues to add capabilities that connect agentic AI applications with context, the vendor needs to address streaming data for agents, according to Dolezal.
Because agents -- unlike static reports and dashboards -- are constantly working independently and acting based on what they know, they perform best when fed the freshest possible information. If they're working based on months-old data, they're making suboptimal decisions and basing actions on those suboptimal decisions.
"Everything [AWS introduced] is oriented around data lakes, warehouses, and static objects, but agents increasingly need to reason over live operational data," Dolezal said.
Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management.
