Your Infrastructure Is Not a Playground: AI Agents for Infra Done Right

AI has transformed many parts of software engineering, but infrastructure has remained cautious, and for good reason. When the blast radius of a mistake is measured in outages, not exceptions, blind automation is not an option.

In this session, I’ll share real-world lessons from deploying AI-driven systems in enterprise and service provider networking using MCP servers. We’ll explore how MCP unlocks new opportunities for safe, scalable automation and housekeeping in network operations, while still respecting the need for guardrails, validation, and human control.

Rather than chasing full autonomy, I’ll show how to design agentic systems that are deliberately constrained: with deterministic and probabilistic guardrails, controlled tool access, and human-in-the-loop decision points where they matter most, like pushing configurations to production.

Through practical examples and a live demo using low-code, multi-agentic worflows with n8n, you’ll learn how to balance innovation with operational safety, and how to build AI systems that infrastructure teams can actually trust. In the end, we don’t want AI to be the reason of a Friday-night outage …