How Booking.com manages Amazon Connect Customer flows as code at scale

This article describes how Booking.com manages Amazon Connect customer flows as code at scale, combining console-first design with infrastructure-as-code promotion.

• Designers compose flows with dynamic references to resolve queues, prompts, and targets at runtime rather than embedding environment-specific ARNs
• Event-driven pipeline detects console changes via CloudTrail and EventBridge, then exports flows as portable templates with placeholders
• Lambda exporter encodes resource names as Base64 identifiers to create environment-agnostic templates stored in S3
• Terraform renders templates using templatefile() to inject environment-appropriate ARNs at deploy time
• Reduced promotion effort from 1,164 engineer-hours to 78 across 90 flows and 242 changes over six months
• Key principle: composition discipline shrinks the problem before automation tackles what remains

The solution enables designers to keep working in the Amazon Connect console while ensuring all production changes ship through CI/CD pipelines as code, eliminating manual JSON editing and cross-environment drift.