AI Agent Governance

Chapter 1. Autonomous Agents as Operational Actors

A note for Early Release readers

With Early Release ebooks, you get books in their earliest form—the author’s raw and unedited content as they write—so you can take advantage of these technologies long before the official release of these titles.

This will be the 1st chapter of the final book. The GitHub repo will be made active later on.

If you’d like to be actively involved in reviewing and commenting on this draft, please reach out to the editor at [e-mail].

For as long as we’ve been writing software, the actors in production environments were predictable. A cron job ran every morning at 2:00 a.m., completed the same task, and exited. An event-driven worker pulled messages off a queue as they arrived. An API with a published contract waited for its first request. The implementation might have been complex enough to need a team of engineers, but the possible paths were defined up front and tested before release.

Agents introduce a new operating model, and governance must be part of it from the start. You give an agent a task, simple or complex, and it decides how to complete it as it goes. The agent may read a production log, decide a critical service is reaching peak capacity, search for a runbook, and restart the service. The next time it reads that same log, it might decide to escalate to a human instead. In agentic systems, the current state of the environment and the prior results influence what comes next. ...