UTurn had sales, leadership, and partnerships team members on the ground in New York last week for the AWS Partner Summit and AWS Summit New York City. We sat through the keynote, worked the floor, and compared notes with the AWS teams building these products. Here is what we took away, and more importantly, what it means for the technical leaders deciding where to place their next bet.

The theme was not subtle. Swami Sivasubramanian, AWS VP of Agentic AI, framed the entire keynote around one idea: agents that compound value over time. The more an agent is used, the more context it accumulates; the more context it has, the better its outcomes; the better the outcomes, the more work you hand it. Over the last six months the conversation has shifted from whether agents work to how fast you can get them into production. Every announcement below is built to close that gap. So is the work we do.

The honest problem: building an agent is easy, running one is not

Anyone can stand up an agent demo. We see it constantly. The hard part is everything after the demo: grounding the agent in your real data, governing what it can touch, keeping it secure, and proving it produces correct answers under production load. AWS reported that tasks performed by agents on Bedrock AgentCore grew 15x in six months. That growth is exactly where pilots go to die if the foundation underneath them is weak. The Summit announcements are an attempt to harden that foundation, and they map directly to the gaps we find in client environments.

AWS Context: the announcement most enterprises should care about first

AWS Context was the launch with the broadest implications. It automatically builds a knowledge graph from your existing data, infers the relationships between your data assets, business rules, and domain knowledge, and makes that available to every agent in your organization. The metadata lands in Iceberg format in S3 Tables, with governance built in so agents only reach data they are cleared to reach.

Here is why it matters: context is what makes an agent's tenth decision better than its first. Without it, agents confidently return answers that are wrong. The catch is that AWS Context is only as good as the data foundation feeding it. A knowledge graph built on top of fragmented, ungoverned, poorly catalogued data will infer the wrong relationships at scale. This is the work UTurn does before the AI conversation even starts. UTurn holds the AWS Data & Analytics Competency precisely because a clean, governed, well-modeled data foundation is the unglamorous prerequisite that determines whether services like AWS Context deliver or disappoint.

AWS Continuum: security has to move at machine speed now

AWS Continuum for code vulnerabilities, launched in gated preview, addresses the full lifecycle of code vulnerabilities at machine speed. It continuously discovers vulnerabilities, validates which are genuinely exploitable, prioritizes by business context, and helps remediate them within guardrails you define. It is model agnostic, every decision is explainable, and every action is auditable. AWS also added Continuum threat modeling, which generates threat models from design documents or source code.

The driver here is uncomfortable but real: attackers are already using specialized security models to find and exploit vulnerabilities faster than human teams can respond. Defending at human speed against machine-speed attacks is a losing position. For our clients in financial services and healthcare, where the cost of a missed vulnerability is measured in regulatory exposure, this is the announcement that justifies a roadmap conversation. The value, though, comes from wiring Continuum into your existing remediation process and guardrails, not from turning it on and walking away.

Amazon Bedrock AgentCore: the production layer got materially stronger

For teams already building on AWS, the AgentCore enhancements were the most immediately useful. The highlights:

The new Managed Knowledge Base handles ingestion, parsing, and retrieval for RAG, with native connectors to S3, SharePoint, Confluence, and Google Drive plus an agentic retriever for complex queries. Web Search on AgentCore grounds agents in current, cited web results without data ever leaving your AWS environment. New optimization capabilities turn production traces into continuous improvement, with A/B testing now generally available. New policy integrations bring Amazon Bedrock Guardrails into AgentCore to evaluate every agent action for prompt injection, harmful content, and sensitive data exposure, with detection signals from Check Point, Zscaler, Rubrik, Netskope, and SentinelOne coming. And the AgentCore harness is now generally available, so you declare what an agent does and AgentCore assembles the orchestration loop, tool execution, memory, and error recovery.

Taken together, AWS just removed a large chunk of the undifferentiated plumbing that slows POC-to-production timelines. That is the same problem UTurn's Realization Factory accelerators were built to solve, and these launches make that path shorter.

AWS Transform and the DevOps Agent: modernization stops being a one-time project

Two announcements addressed the reality that code starts aging the moment it ships. AWS Transform continuous modernization is an always-on capability that finds tech debt, fixes it, validates the fix, and learns from each transformation, plugging into CodePipeline, Jenkins, GitHub Actions, and GitLab. AWS noted Transform has already eliminated over 1.6 million hours of manual effort for customers including BMW Group and Experian. The AWS DevOps Agent added release management: release readiness reviews and autonomous, change-specific release testing so your pipeline can keep pace with agent-accelerated development.

The strategic shift worth internalizing: when agents write code ten times faster, your bottleneck moves downstream to release safety and modernization. Speed without a release discipline just ships regressions faster. This is squarely where UTurn's Migration & Modernization and DevOps Competencies live, and where an embedded partner who operates as an extension of your team beats a project shop that hands off code and disappears.

Amazon Quick, Kiro for iOS, and the rest

A few launches round out the picture. Amazon Quick added no-code autonomous agents that run in the background, a prioritized activity feed across email, messaging, calendar, and tasks, and 16 new integrations including Adobe, Moody's, and Snowflake. Kiro, the AWS coding agent, is now available as a native iOS app in gated preview, with always-on cloud sessions you can steer from your phone. Amazon S3 annotations let you attach up to 1 GB of queryable context directly to objects. And AWS WAF added a capability for content owners to meter and charge AI bots for access, a quiet but interesting signal about how the economics of the agentic web are forming.

The proof point AWS put on stage

The clearest validation came from a customer, not a product. Southwest Airlines announced it is moving from a largely on-premises environment to a cloud-based, AI- and agent-enabled architecture on AWS by 2028, with more than 2,700 developers already using Kiro and an AI-Driven Development Lifecycle approach where agents move work forward while engineers guide and validate. That is the model in practice: a regulated, operationally complex enterprise treating agents as production infrastructure, not experiments.

What to do with this

The announcements all point in one direction. The agentic stack on AWS is maturing fast, and the differentiator in 2026 is no longer access to the tools; it is whether your data foundation, security posture, and delivery pipeline are ready to absorb them. Most of the value sits in the integration work between these services and your environment, and that is exactly the work UTurn does in production every day.

If your team is evaluating where AI agents fit in the next twelve months, UTurn's AI Readiness Assessment is built to pressure-test your data foundation, security guardrails, and delivery pipeline against the agentic patterns AWS just put on stage, so your pilots have somewhere solid to land.