I’ve Used 100+ ChatGPT Prompts in DevOps — Here’s What I Stopped Doing

After using ChatGPT for everything in my DevOps work — from CI builds to incident triage — I realized certain prompts weren’t just useless… they were slowing me down.

Not a Member?? Click Here

Look, I’ve been that DevOps engineer who thought AI was going to solve everything. Six months ago, I was throwing every infrastructure problem at ChatGPT like it was some kind of digital oracle. After burning through 100+ prompts across real production workflows — some brilliant, others catastrophically wrong — I learned what NOT to do the hard way.

Here’s what I stopped doing, and why it made all the difference.

1. Prompting Without Role or Context

What I used to do:

"Help me write pipeline config."

But: ChatGPT gave me generic config and missed my specific AWS/GitOps context → huge rework. I got a beautiful Jenkins pipeline when I was using GitHub Actions. Spent 4 hours translating YAML that should’ve taken 30 minutes.

Therefore: I always start prompts with: “You are a senior DevOps engineer building pipelines on AWS with GitOps” so the outputs are relevant.