Why Using AI Less Made Me More Valuable
I switched from full-stack to SRE and stopped leaning on GitHub Copilot constantly. My understanding—and my job security—actually went up.
Three months ago, I was using GitHub Copilot agent mode for almost everything: implementing features, debugging, writing commits, even moving repos. It felt productive. I was shipping fast.
Then I switched into SRE, and something flipped.
The Old Way Felt Productive (But It Wasn't)
As a full-stack developer, I'd ask the agent to implement something, check for errors, publish branches—basically let it handle the entire development workflow. It was quick. It rarely broke things. But here's the thing: I didn't actually understand why any of it worked.
When you're building dashboards and fixing bugs in code you wrote, AI can genuinely replace you. It knows the patterns. It knows the domain. It can do the job.
SRE Forced Me to Think Differently
In SRE, the work is different. I'm doing knowledge transfer sessions, reading Confluence docs, understanding how patterns are implemented, looking at architecture design. The why matters here in a way it didn't before.
I still use AI—to get AWS service examples, to understand concepts behind actions. But the frequency dropped drastically. I'm not asking it to implement entire workflows anymore. I'm asking it to explain.
And my understanding went up a lot.
The Real Test: When AI Couldn't Help
Last week I hit an onboarding story: deploy and monitor an Artemis alarm in AWS. The docs were outdated. Links were broken. Steps didn't match reality.
I followed the Terraform steps in GitLab, but the alarm never actually deployed. The documentation said it should. The CI/CD pipelines looked right.
I dug into the repo, worked through it with ChatGPT—and found a folder mismatch and a version typo. That's fine. But I was spending hours on something that should've been faster.
Then I asked a teammate a direct yes-or-no question: "Did you try local execution instead?"
He said yes. I pivoted immediately, read how local execution works, and unblocked myself in minutes.
AI couldn't give me that institutional knowledge. It couldn't replace a quick human conversation.
The Actual Difference
When you don't know the why, AI can replace you pretty easily. When the why is the entire job, it can't.
As a junior full-stack dev, Copilot was doing the exact work I was doing. In SRE, I'm doing work that requires system understanding, knowing how things are connected, and most importantly—knowing who to ask.
I'm more valuable now not because I'm smarter. I'm more valuable because I can't be outsourced to a tool.
If you're using AI constantly right now and shipping fast, we were in the same place. That's okay. But the work that actually protects your career is the work where you understand the why. The rest is just automation waiting to happen.