I wanted to share a recent blog post that explores a topic a lot of us are actively thinking about right now: how agentic AI can interact with automation systems in a way that is secure, auditable, and aligned with existing operational controls.
The post introduces the Model Context Protocol (MCP) server for Red Hat Ansible Automation Platform and explains how it can be used as a control plane between AI agents and automation. Rather than letting an LLM or assistant talk directly to infrastructure or APIs, MCP provides a structured interface that exposes only approved automation actions, while still relying on Ansible’s existing RBAC, credentials, and execution model.
What I found especially useful in this write-up is that it does not treat agentic AI as a replacement for automation, but as a new interaction layer. The AI agent can reason, suggest, and request actions, but Ansible remains the system of record that actually performs changes. That separation helps preserve determinism, governance, and trust, which are often the first concerns raised when AI is introduced into operations workflows.
The blog walks through:
- Why MCP is useful for bridging AI agents and enterprise automation
- How MCP integrates with Ansible Automation Platform rather than bypassing it
- Example interaction patterns where AI assists operators without taking uncontrolled action
I put together a video walkthrough using the AI client Cursor as an example of how this can work: https://youtu.be/EidwVmZQkGM?si=IMK9olAd1YworDpd