On Linux VM, if a program is given root privilege and it just execute whatever command ChatGPT spit out, then you'll got hell a problem solver/havoc wreaker.
The hit rate of AI is still in that phase of being right almost all the time and being spectacularly wrong 2% of the time. So actually pretty close to a good DevOps engineer.