>LLMs are great in both areas.

Nuance. LLMs are just going to report that they cant SSH to an endpoint, after delivering your vibeconfig, and throw it back to you to resolve connectivity. Your velocity with LLMs will stall at break fix every time.

>AI makes AMAZINGLY good network operations triage agents, since they are able to immediately process so many signals.

I have seen a lot of tokens spent on solutions that could have just been grafana.

I'm a network engineer at a F500 and am no stranger to automation. We use a lot of Ansible and Terraform to manage both on-prem and AWS network infrastructure. Lots of Juniper and Cloudflare products, and we use NetCM, Netbox, Prometheus, and Grafana. But we've been playing with agents since late 2024, and our team has come a long way since then.

It was already easy enough and straightforward to deploy a network because we had built so many CLI tools to handle what we needed, but it still required a bit of a human touch to validate outputs and feed those to different tools. Thanks to skills with helper scripts, we're pretty close to one click deployments these days. So much of our maintenances or operations can even be handled from our phones. We can just tell an agent that there's a new version, or AMI, and we can reliably trust them to safely update the fleet from end to end without causing service disruptions. When customers need updates made, agents draft the PRs and I just review, and they deploy after I approve.

I would argue that most of what makes this possible though isn't LLMs themselves, but having invested in robust network design, consistent standards (snowflake setups or configs will cause you problems), proper observability, and detailed docs/runbooks. While I have my doubts that someone will ever be able to vibe code that knowledge and experience, you can certainly use agents to amplify an already strong engineering foundation.