AWS Makes Agentic Push At AWS Summit New York City 2025

AWS Summit New York City is a smaller cousin to the re:Invent mega-conference scheduled for the late fall in Las Vegas, so most big announcements tend to wait until that event, but agentic AI won’t wait, despite its relatively modest take-up to date. AWS’s key announcement was Bedrock AgentCore, intended to redress a perceived market imbalance between AWS and its primary competitors in the agentic AI space. In classic AWS fashion, it is pursuing a developer-friendly, partner-dependent vision for agentic AI leveraging key global systems integrator (GSI) partners.

AWS has also announced the ability to perform customization of Amazon Nova foundation models with SageMaker AI across stages of model training. AWS furthermore announced a number of promotional offerings, including the AWS AI League for upskilling and credits for its Free Tier program for users of its AI solutions. Leading GSI partners were heavily involved in AWS’s analyst program.

While AWS presented its own use of agentic capabilities in its own business, one of the key messages from the event — not just from AWS but from its GSI partners — is that agentic is not for everyone (or everything). Yet to not pursue it, despite its incomplete and imperfect nature, leads to risks of falling behind.

Additional announcements of interest:

Amazon S3 Vectors, a new cloud-based object store providing native support for storing and querying vectors, a significant offering for the generative AI era.
New capabilities for Amazon SageMaker, including Amazon QuickSight integration for dashboard creation, governance, and sharing.
New capabilities for Amazon S3 unstructured data integration, intended for cataloging documents and media files, and automatic data onboarding for Lakehouse. The goal: eliminating data silos by unifying structured and unstructured data management visualization and governance.
Acceleration of safe software releases with new built-in blue/green deployments in Amazon ECS.
Amazon EKS support for very large AI/ML workloads, with support for 100,000 nodes per cluster.

These offerings may appear to be incremental at a moment when one blockbuster genAI announcement about model advances is quickly followed by another. But for AWS customers, they are key elements in enabling adoption and rollout of genAI workloads, leveraging their data gravity within AWS and services already in place to create new genAI options. For additional public cloud insights, read Forrester’s report, Buyer’s Guide: Public Cloud Platforms, 2025.