From Runbooks to Production: AI-Orchestrated Infrastructure at Enterprise Scale
Date:
How do you turn a two-week, multi-account infrastructure runbook into a one-hour conversation? This session presented a re-imagined developer portal experience built on Strands Agents SOPs: architecture decisions, security constraints, and deployment steps codified once as machine-readable runbooks (RFC 2119 keywords, parameterized inputs, progress tracking), executed by an agent that generates Terraform and application code, monitors deployments, and adapts to failures — with a human setting parameters, reviewing the plan, and gating production. The live demonstration took a web application from description to production across three AWS accounts: ~150 resources, 6 Terraform modules (VPC, ALB, CloudFront, ECS, RDS, Cognito), in under an hour. Presented at the AWS AI SDLC Workshop in Vienna.
The target architecture the agent bootstraps: a three-tier web application deployed by a single CI/CD pipeline across dev, staging, and production accounts, with a manual approval gate before production.