Amazon's New AI Automates Tech Diagrams So Devs Work Faster
Outdated diagrams cause delays — this AI updates them every time code changes.
Amazon Web Services (AWS) has introduced AgentCore, a service that lets companies build AI "agents" — software that can take actions on its own. One of its first practical uses is solving a boring but expensive problem: keeping architecture diagrams up to date. Architecture diagrams are the maps that show how pieces of a software system connect. Most teams draw them by hand, and they become outdated within weeks as programmers change code. That leads to confusion, slow bug fixes, and problems when regulators ask to see how a system works.
What makes this new tool different is that it runs automatically. When a developer saves a code change to AWS CodeCommit (a place to store code), the system wakes up. An agent built with AgentCore reads the new code, identifies the big pieces, and creates new diagrams in formats like SVG or Mermaid. Those diagrams are then stored in Amazon S3 (simple cloud storage) and made searchable with Amazon Bedrock Knowledge Bases, which lets engineers ask questions about their system in plain English. This happens continuously, so the map always matches the real system.
A global interdealer broker — a company that helps large banks trade — has been using this in production since early 2026 to document its electronic trading platform. The agent runs inside AWS CodePipeline, the company's automated release process, so every time code is pushed, the diagrams refresh. New developers no longer have to reverse-engineer old code, and auditors see up-to-date records. The payoff is measurable: less downtime, fewer surprises, and more time spent on building things.
But there's a catch. This only works if your software lives on AWS, and you still need engineers to set up the agent and make sure it captures important details. It's not magic — it's automation mixed with human judgment. Still, if you work at any company that builds digital products, more accurate "living documentation" means your services become more reliable, updates ship faster, and the time spent on overhead drops.
- Amazon's AgentCore uses AI to automatically read computer code and turn it into diagrams that show how a system is built.
- The result is stored in a searchable library, so developers can ask natural questions like "which services talk to the payment system?" without reading thousands of lines of code.
- A global financial trading firm has had this running in production since early 2026, proving it works in demanding environments where accuracy matters.
Why It Matters
Accurate software diagrams mean fewer outages, quicker fixes, faster hiring, and lower compliance risk for any business.