Infrastructure management has evolved from manual CLI to AI agents, but legacy systems, brownfield environments, and resistance to change still pose significant challenges.
The Model Context Protocol (MCP) acts as a universal standard, like USB-C, for AI agents to discover and use tools without extensive pre-training.
Low-code platforms like n8n enable rapid prototyping of AI agents by providing a drag-and-drop interface that balances probabilistic AI with deterministic nodes.
A live demonstration shows how to quickly construct a networking AI agent using n8n, an LLM, and an MCP client to query device inventory.
A two-agent system uses a planning agent to create a dry run and a human-in-the-loop approval via Slack before a commit agent applies infrastructure changes.
An automated workflow uses AI agents to run scheduled compliance reports against infrastructure, generate markdown summaries, and create GitHub issues for non-compliant findings.
Follow key recommendations such as taking baby steps, treating agents like junior engineers with specific personas, and keeping workflows simple for safe and effective deployment.
Jobs that call for the skills explored in this talk.