Agentic coding at Amazon - Lessons from Amazon and our customers

Learn practical lessons from Amazon and AWS on effectively scaling agentic coding tools across an organization, based on real-world rollout experiences.

Kiro Software

Overview

AI coding assistants are transforming software development, but adopting them effectively requires more than just giving developers access to new tools. Drawing on real-world experience from rollout to Amazon’s internal developers and AWS’ external customers, this session shares practical lessons on what actually works when scaling agentic coding across an organization

Tech stack

Kiro Software

An agentic software development service built on Amazon Bedrock that uses spec-driven workflows to turn natural language prompts into production-ready code, tests, and pull requests.

Kiro is an AI-powered development platform that brings engineering rigor to agentic workflows (available via IDE, CLI, and Kiro Web). Built on Amazon Bedrock, it moves beyond basic autocomplete by using a spec-driven methodology: the system first generates structured requirements, design documents, and step-by-step task plans before writing any code. Developers can steer the agent using custom steering files, automate routine tasks with event-triggered agent hooks, and integrate external tools via Model Context Protocol (MCP) servers. By coordinating changes across multiple repositories simultaneously, Kiro automates complex, multi-step engineering tasks and delivers fully verified pull requests directly to your repository.

https://kiro.dev

View projects