Itential outlines AI steps for network engineers

Itential published a podcast-based brief reporting that roughly 70% of networks remain without automation, recommends introducing Artificial Intelligence (AI) through read-only workflows, and highlights the Model Context Protocol (MCP) as a way to connect AI agents to network tools.

Research overview

The summary derives from a podcast conversation that referenced a Network Automation Forum survey showing about 70% of networks are not meaningfully automated and noted growing divergence between adopters and nonadopters.

The discussion framed AI adoption for network engineering as an incremental progression built on existing automation foundations rather than an immediate switch to full configuration control.

Key findings

The author reported that a minority of environments have automated foundations and that those teams are beginning to layer AI onto automation, widening a capability gap with less automated teams.

The post attributed low automation uptake to three broad causes: fragmented tooling and data models, a cultural expectation that engineers operate as programmers, and limited organizational mandates to convert manual practices into automated processes.

Technical breakdown

The MCP is presented as a client-server mechanism that exposes diverse resources—API calls, scripts, database queries, or Command-Line Interface (CLI) output—to AI agents as callable tools, implemented over JSON-RPC, and the author cited roughly 17,000 MCP servers in use.

The blog offered three MCP examples for network teams: an RFC-backed server for standards-aware configuration assistance, an agent-connected server that generates and evaluates interface health commands, and NetBox/Nautobot integrations recommended as initial test targets in nonproduction environments.

Operational impact

The author recommended starting AI integration in a read-only mode to build trust, using agents to summarize logs, compare intended and deployed configuration, perform compliance checks, and generate dynamic Border Gateway Protocol (BGP) health tests before any automated changes are applied.

The recommended approach includes using demo or sandbox environments for initial experiments, obtaining Application Programming Interface (API) keys for test instances, and considering local model options to address privacy or cost concerns while avoiding production systems.

Leadership perspective

The post framed agent deployment as an organizational design issue, noting that large-scale agent use requires defined permissions, guardrails, and a plan for encoding senior engineering knowledge to prevent unsafe actions such as exposing management interfaces or changing secrets.

The author argued that early automation adopters were not displaced and reported that building agents can preserve institutional knowledge and create internal capabilities that leadership tends to retain and extend rather than remove.

This “Blog Signals brief” summarizes the vendor blog for enterprise decision-makers and highlights practical adoption steps, operational safeguards, and the MCP integration path; this “Blog Signals brief” is a fact-based summary of the vendor blog.