The Big Picture: AWS’s Three Core Themes

Agentic AI as a First-Class Workload
Autonomous, multi-step AI systems built with governance, observability, and safe action-taking.
AI-Driven Modernization at Scale
AWS Transform compresses multi-year modernization timelines into months.
Operations for the AI Age
CloudWatch, Application Signals, and MCP servers deliver unified, AI-powered observability.

Agentic AI: Building Autonomous Systems at Scale

Amazon Bedrock AgentCore: Now with Guardrails and Monitoring

AgentCore evolves into a platform for enterprise-grade multi-step AI agents, introducing:

• Quality Evaluations
Monitors agent behavior, detects drift, and measures compliance with business rules.

• Policy Controls
Tightly scoped permissions to ensure agents operate safely inside predefined boundaries.

• Experience-Based Learning
Agents improve from real-world interactions while capturing governance signals.

Together, these enable enterprises to safely take AI agents into production at scale.

Lambda Durable Functions: Stateful Workflows Without Paying for Idle

A new model that lets you run multi-step, long-running workflows using simple code — with no state stores, no Step Functions overhead, and no cost while waiting.

What it unlocks:
• Agentic workflows with pauses & resumptions
• Long-running SaaS onboarding/approval flows
• Distributed transactions & Saga patterns
• Complex operations without orchestration sprawl

This is one of the most powerful serverless updates in years.

AI-Powered Modernization: Code → Cloud-Native Architecture

AWS Transform: The Modernization Superpower

Transform is now AWS’s signature generative AI modernization suite, covering Windows, mainframes, and custom enterprise codebases.

1. Windows Modernization

Transforms:

C#, VB.NET, ASP.NET apps
UI frameworks
Database schemas & stored procedures
Infra & deployment configs

➡ Up to 5× faster modernization, turning 18-month refactors into 3–4 months.

2. Mainframe Reimagine

Automates:

COBOL/PL/I → modern cloud-native code
Functional test generation
Infrastructure-as-code output
Architecture modernization in months

A breakthrough for organizations stuck on mainframes.

3. Custom Enterprise-Wide Transform

Learns your patterns, refactors your entire repo, normalizes architecture, logging, frameworks, and structure.

Modernization is no longer manual—it's generative.

Intelligent Cloud Operations: The New Observability Stack

CloudWatch: Now an AI-Driven Operations Intelligence Platform

Generative AI Observability for Agents

Native support for:

Latency, token usage, error metrics
AgentCore, LangChain, LangGraph, CrewAI
Trace-level agent decision visibility

CloudWatch Investigations

AI-driven root cause analysis using AWS’s internal incident methodology.

Application Map (Zero Instrumentation)

Auto-discovers service topology and dependencies. No agents. No libraries. No overhead.

Cross-Account & Cross-Region Centralization

Centralized logs, metrics, and database insights across orgs and hybrid setups.

CloudTrail + CloudWatch Logs Upgrades

Aggregated data events for reduced noise and cost
OCSF-native normalization for security
Apache Iceberg export for analytics

Security and compliance teams finally get a unified logging strategy.

Compute, Containers, and Infra

Lambda Managed Instances

A new hybrid model:
• Serverless experience
• EC2-level control
• GPU and custom hardware options
• Savings Plans compatibility

Blurs the boundaries between Lambda and EC2.

Amazon EKS: Toward Fully Managed Kubernetes

More opinionated defaults
Automated infra orchestration
Lower operational burden
AWS-grade reliability baked in

EKS is becoming the enterprise-ready Kubernetes baseline.

Database & Data Services

Database Savings Plans + RDS Enhancements

Flexible long-term DB discounts
SQL Server Developer Edition on RDS
New M7i/R7i instances
Storage scaling up to 256 TiB

Cost optimization takes center stage.

Amazon S3 Tables

SQL-friendly structured data in S3
Intelligent-tiering
Replication & versioning
Storage Lens performance visibility

S3 becomes a first-class analytics substrate.

Developer Experience & MCP Ecosystem

IAM Policy Autopilot (Open-Source MCP Server)

AI agents can now generate least-privilege IAM policies by analyzing your code.
Huge for DevOps and security automation.

CloudWatch MCP Servers

Agents can query metrics, logs, alarms, traces—directly.

This is the future of AI-driven cloud operations.

AI-Augmented AWS Support Plans

Proactive monitoring, cost optimization, and guided troubleshooting built in.

What This Means for Builders

1. Agentic AI is no longer optional
Bake AgentCore into your platform strategy.

2. Observability is now a competitive moat
Centralized, AI-powered ops will define high-performing teams.

3. Modernization is becoming AI-first
Transform is the new benchmark—manual refactors won’t scale.

4. MCP is the interface layer for AI agents
Expose your platform APIs through MCP now.

5. Serverless + stateful is the future
Durable Functions unlocks an entirely new design space.

Final Thoughts

AWS re:Invent 2025 signals the beginning of autonomous, observable, AI-driven cloud operations. The shift is clear: AWS is laying the primitives for the agentic era, and the next generation of platforms will be built on top of these capabilities.

What excites you most from re:Invent 2025?
Which services are you planning to adopt?

Drop your thoughts below.