Agile for Agents: Managing Robots The Way We Manage Humans

After a few months of building software with ChatGPT and Claude, I discovered that code generation wasn’t the hard part anymore. Coordination was.

AI agents forgot previous decisions, repeated work that was already finished, and struggled to pick up where another session left off. After running into these problems repeatedly on real projects, I began adapting techniques from Scrum, Kanban, and traditional project management to create the Agent Management System (AMS), a simple collection of markdown files that provides project memory, documentation, and handoffs across different AI tools.

In this session we'll cover:

- The Agent Handoff Protocol: how agents start, continue, and conclude work
- The AMS directory structure and the project artifacts it uses
- Using specialized agent roles such as coder, librarian, and tester
- Practical integrations with Claude Code and other AI coding tools
- Lessons learned from real Drupal and open-source projects

Examples throughout the session are drawn from Drupal projects, but the techniques work with any software project and most other forms of knowledge work.

Whether you’re just starting to use AI coding tools or already managing a fleet of agents, you’ll leave with a working framework you can drop into your next project.