TL;DR — Amazon used the AWS Summit NYC 2026 stage to launch eight new agentic AI products in a single keynote, the largest single-day agent product release in AWS history. The lineup spans AI-native security (AWS Continuum), enterprise knowledge graphs (AWS Context), autonomous desktop agents (Amazon Quick), mobile orchestration (Kiro app), devops release management (AWS DevOps Agent), continuous modernization (AWS Transform), and a significant Bedrock AgentCore expansion with web search and managed knowledge bases. Southwest Airlines was named a flagship customer with 2,700+ developers now using Kiro.
On June 17, 2026, Swami Sivasubramanian — AWS's VP of Agentic AI — took the stage at the Javits Center in New York and delivered what amounts to Amazon's most aggressive agentic AI product salvo to date. In a single keynote, AWS launched or significantly upgraded eight products across the agent lifecycle: from security scanning at machine speed to autonomous desktop agents that claim to "reclaim your time."
The throughline was clear: Amazon believes the window for enterprise AI adoption is narrowing, and it's betting that comprehensive infrastructure — not just models — will win the agentic platform war.
"Over the last six months, there has been a seismic shift as companies have started to move from talking about agents to putting them to work," Sivasubramanian said. "Agents are building apps, securing systems, answering complex customer questions, and making decisions autonomously." (Source: About Amazon — AWS Summit New York 2026: New AI agent innovations)
The 8 Announcements: A Product-by-Product Breakdown
1. AWS Continuum — AI-Native Security at Machine Speed
The most strategically significant launch of the day. AWS Continuum for code vulnerabilities is an AI-native security service that handles the full lifecycle of vulnerability management — discovery, validation, prioritization, and remediation — continuously, not in batch scans.
The timing is no accident. Sivasubramanian explicitly referenced the emergence of specialized security models like Claude Mythos, noting that "the use of these models by attackers and defenders has greatly accelerated the ability to find and exploit vulnerabilities." His message: traditional CI/CD pipeline security can't keep up with AI-speed attacks. You need agents scanning continuously.
Continuum is model-agnostic by design — it routes different vulnerability types to the best model for the job and integrates new models as they emerge. At every stage, users get full visibility into what Continuum is doing, why it suggested a specific action, and what the rollback plan looks like. This is the "guardrails within guardrails" approach — autonomous enough to move at machine speed, transparent enough that security teams don't lose control.
(Source: About Amazon — AWS Summit New York 2026)
2. AWS Context — The Knowledge Graph That Gives Agents Ground Truth
Every enterprise agent builder has hit the same wall: the agent is only as good as the data it can access. AWS Context is Amazon's answer — a comprehensive knowledge graph designed to give agents access to enterprise-wide context without manual integration work.
Context learns, connects, and works across all of a company's data (structured and unstructured), enabling agents to deliver grounded answers rather than hallucinating from incomplete context. It's the enterprise version of what makes consumer AI assistants useful — knowing where you left off, what tool you prefer, and what data is relevant to the current task.
"With Context, agents will know where to get the information they need, provide the right answer or take the right next step, faster," Sivasubramanian explained. This directly addresses a finding from the recently published Confluent 2026 Data Streaming Report, which found that data governance and quality are the top obstacles IT leaders cite for moving agentic applications from pilot to production. (Source: Confluent 2026 Data Streaming Report)
3. Amazon Quick — Autonomous Agents for the Desktop
Quick graduated from assistant to autonomous agent. The new Quick agents "work on your behalf across all your web and desktop apps taking high-quality, autonomous actions." Amazon's pitch: Quick handles the busywork — booking meetings, filling forms, pulling reports — while you focus on higher-value work.
This puts AWS in direct competition with the emerging "computer use" agent category pioneered by Anthropic's Claude and extended by companies like Adept. The differentiation is AWS-scale infrastructure behind it, plus deep integration with the rest of the AWS agent ecosystem.
4. Kiro Mobile App — Agent Orchestration on the Go
Kiro, AWS's agentic coding service, now has an iOS and Android mobile app. Developers can approve pull requests, review automations, and manage agent tasks from their phone. This isn't just a convenience feature — it's a signal that AWS sees agent orchestration as a continuous activity, not something that stops when you leave your desk.
Southwest Airlines was revealed to have 2,700+ developers already using Kiro to build features, automate testing, and generate cloud infrastructure for the modernization of Southwest.com. Southwest is also adopting what AWS calls an "AI-Driven Development Lifecycle (AI-DLC)" where AI agents move development forward while engineering teams guide and validate outcomes.
5. AWS DevOps Agent — Release Management
AWS DevOps Agent now includes Release Management capabilities. The pitch: "your agents don't just write code, they help you ship safely and reliably." This extends the agent's role from code generation through deployment, adding guardrails for production releases. It's another step toward end-to-end agent-driven software delivery pipelines.
6. AWS Transform — Continuous Modernization
AWS Transform's new continuous modernization capability is designed to "keep you ahead of tech debt." The service continuously analyzes codebases and infrastructure, proposing and implementing modernization changes. In a world where agents are generating more code than ever, having an agent that also manages the resulting technical debt is a logical — if ambitious — addition to the portfolio.
7. Amazon Bedrock AgentCore — Web Search, Managed KBs, and Harness GA
The AgentCore platform received a significant upgrade with three new capabilities:
Built-in Web Search with zero data egress charges. Agents can now fetch real-time information from the web without leaving the Bedrock environment. This is a direct response to the "grounding problem" — agents that can't access current information quickly become outdated.
Managed Knowledge Base — a fully managed service handling ingestion, parsing, chunking, embedding, and storage for enterprise data. It replaces the DIY pipeline most teams currently build themselves.
AgentCore Harness (General Availability) — a declarative way to build agents. "You simply declare what your agent does (the model it uses, the tools it calls, and instructions to follow) and AgentCore handles the rest," per the announcement. Under the hood, it assembles the orchestration loop, tool execution, memory management, context handling, and error recovery.
Agent Optimization — new tools to benchmark and improve agent accuracy, latency, and cost.
The Bedrock enhancements also include integration with security partners like CrowdStrike and SentinelOne, and Rubrik announced an upcoming integration with Bedrock AgentCore to secure AI agents.
(Source: AWS Machine Learning Blog — New in Amazon Bedrock AgentCore)
8. Southwest Airlines — The Flagship Enterprise Win
Southwest Airlines was named as a flagship customer. The airline is partnering with AWS as its preferred cloud provider to modernize from a largely on-premises environment to a cloud-based, AI- and agent-enabled architecture by 2028. Southwest will transition its entire technology foundation to AWS, with 2,700+ developers already actively using Kiro and expanding into Amazon Quick.
The Southwest deal validates the thesis that enterprise agent adoption isn't about a single tool — it's about an entire stack, from security to knowledge management to orchestration.
(Source: AWS Press Center — Southwest Airlines Partners with AWS)
Analysis: The Compounding Momentum Thesis
Sivasubramanian's keynote introduced a concept he called "compounding momentum" — a framework for understanding how enterprise AI adoption accelerates. The logic is elegantly simple:
- More agent interactions → more context gathered
- More context → better outcomes
- Better outcomes → increased trust
- Increased trust → more work handed off to agents
This flywheel, he argued, "widens the gap between companies that embrace AI and those who do not." It's a clear competitive posture: AWS wants to be the platform that makes this flywheel spin fastest.
The 8-product release cadence also suggests AWS is betting that breadth matters more than any single breakthrough feature. Rather than one killer product, they're delivering a complete agent lifecycle platform — build, secure, ground, orchestrate, deploy, modernize — and betting that enterprises will prefer the integrated suite over point solutions from competitors.
Competitive Landscape: Where AWS Stands
AWS's agentic AI push lands in an increasingly crowded market:
| Competitor | Agentic Play | AWS Counter |
|---|---|---|
| Microsoft | Copilot + Azure AI Agent Service + KPMG partnership | Bedrock AgentCore (multi-model), Kiro (developer-focused) |
| Google Cloud | Vertex AI Agent Builder, Gemini agents | Broader model support, Trainium infrastructure |
| Anthropic | Claude + computer use | Model-agnostic Bedrock (includes Claude), Continuum security |
| OpenAI | GPT models on Azure/Bedrock, custom agents | AgentCore harness for OpenAI models on AWS |
The AI Business analysis noted that AWS's tools "trail rivals" in some areas but "respond to real problems" — enterprise grounding, security at scale, and the messy reality of legacy infrastructure modernization. (Source: AI Business — AWS's New Agentic Tools Trail Rivals, but Respond to Real Problems)
FAQ
Q: What's the most important announcement for security teams?
AWS Continuum is the standout. It's AWS's first dedicated AI-native security service that handles vulnerability management end-to-end at machine speed, model-agnostic, with full transparency into every action. If it delivers on the promise, it could fundamentally change how security teams think about vulnerability management cadence.
Q: Is AWS Context just another vector database?
No. AWS Context is positioned as a comprehensive knowledge graph — it learns relationships between data across the enterprise, not just embeds documents. It's more like a living map of the company's data that agents navigate, rather than a retrieval store they query.
Q: How does this compare to the 2025 AWS Summit announcements?
The 2025 Summit introduced Bedrock AgentCore as a concept. The 2026 Summit is about making it real — Harness is now GA, Managed Knowledge Base is a product, web search is built-in, and there's a whole ecosystem of complementary services (Continuum, Context, Quick, Transform) that weren't on the radar last year.
Q: Will this accelerate enterprise agent adoption?
The Southwest Airlines deal — 2,700 developers on Kiro, full cloud migration by 2028 — is a strong signal. But the Confluent report's finding that 32% of organizations are running agentic AI in production (up from 29% in 2025) suggests the market is moving, but not at breakneck speed. AWS's bet is that the integrated platform approach removes enough friction to accelerate that number.
Q: What's missing?
Noticeably absent: pricing details for Continuum and Context, concrete benchmarks for Quick's autonomy vs. competitors, and any mention of multi-agent coordination beyond the orchestration layer. These will likely come at re:Invent 2026.
Further Reading
- About Amazon — AWS Summit New York 2026: New AI agent innovations
- AWS Machine Learning Blog — New in Amazon Bedrock AgentCore
- AWS Press Center — Southwest Airlines Partners with AWS
- Confluent 2026 Data Streaming Report
- AI Business — AWS's New Agentic Tools Trail Rivals, but Respond to Real Problems
- GeekWire — Amazon unveils new AI agents trying to thread the needle between autonomy and human control
- Introducing Amazon Bedrock Managed Knowledge Base
Cet article a ete initialement publie sur The Agent Report.
Top comments (0)