Can AI agents diagnose cloud outages on their own?

We tried an experiment with Cloud World Model today.

We gave Grok access to our API and asked:

“Try the GitHub-Inspired Cascading Retry Storm scenario and tell me what happened. Why did the outage occur?”

Grok loaded the scenario, ran the simulation, watched the failure unfold, and came back with an RCA.

It identified a hidden sidecar concurrency limit combined with zero-backoff retries that created 9.5× request amplification across the auth/token path.

What interested me wasn’t whether Grok could summarize an outage. It was that an external AI agent could run the experiment itself, observe the consequences, and reason about the root cause.

That raises a bigger question:

If agents can safely experiment with infrastructure before touching the real cloud, how much of reliability engineering could eventually become agent-driven?

Curious how other builders think about this. Where would you want a human in the loop?