From 'Is This Cheating?' to Building AI Agents: How AI Transformed My SRE Career

When ChatGPT was released, I was hooked immediately. The ability to ask a question and get a detailed, useful answer in seconds? To generate scripts, troubleshoot errors, and brainstorm solutions faster than ever? It felt like a superpower. I was using it for everything I could think of.

But I'd be lying if I said it didn't also feel a little bit like cheating.

The Early Days: Excitement Meets Uncertainty

At the time, nobody really knew what to do with it. The tech was incredible, but the unknowns were just as big. Security concerns, data privacy, intellectual property questions. Could company data leak through prompts? Were responses accurate enough to trust? What happens when proprietary code ends up in a training dataset?

My employer at the time discouraged using it, and honestly, I understood why. We didn't have answers to those questions yet. The industry was still figuring out what AI meant for day-to-day engineering work. So I kept my usage mostly personal and experimental, learning what it could do without putting anything sensitive at risk.

It was frustrating though. I could see the potential clear as day, but the guardrails weren't in place yet to use it responsibly at work.

Fast Forward: Everything Changed

The AI landscape today looks nothing like those early ChatGPT days. Agents, sub-agents, multi-step reasoning, tool use, code generation that actually understands context. The pace of evolution has been wild. What felt like science fiction two years ago is now table stakes.

More importantly, the enterprise side caught up. Companies figured out how to deploy AI securely. Data boundaries, private instances, compliance frameworks. The security concerns that held everyone back early on have largely been addressed, at least for organizations willing to invest in doing it right.

My Current Position: AI Is Not Just Allowed, It's Encouraged

I've moved on to different employers since those early days, and at my current position the culture around AI is completely different. We're highly encouraged to use it. We have enterprise plans with no limits. There's no side-eye for using AI to get work done. It's expected.

That shift alone has been massive. Going from "don't use this, we don't understand the risks" to "use this as much as you can, here's an enterprise license" felt like going from dial-up to fiber.

GitHub Copilot is a core part of my daily workflow. It's integrated directly into my IDE, and it accelerates everything from writing Terraform modules to building scripts to knocking out boilerplate code that used to eat up hours. I keep my work tools and personal tools completely separate for obvious reasons, so Copilot is what I use professionally and it delivers.

Building Agents That Do the Heavy Lifting

The real game-changer has been building agents. Not just using AI as a chat assistant, but creating automated workflows that handle entire processes end to end.

I've built agents to:

• Gather and analyze data from multiple sources, correlate it, and surface what matters
• Generate reports that used to take hours of manual work
• Detect patterns across infrastructure metrics that are tedious to spot manually
• Automate repetitive operational tasks that used to eat up large chunks of my week

These aren't simple scripts. They're intelligent workflows that adapt based on the data they receive. The difference between a bash script and an AI agent is the difference between a static dashboard and one that tells you what to look at and why.

Eliminating the Mundane

Every SRE knows the feeling. You have a list of high-impact projects you want to tackle, but your day gets consumed by repetitive, low-value tasks. Pulling metrics, formatting reports, triaging alerts that turn out to be noise, updating documentation. It's not that these tasks don't matter. They do. But they don't require deep thinking, and they steal time from the work that does.

AI handles that now. The mundane, pattern-based, repetitive work that I used to grind through is largely automated. I still review outputs and validate results because trusting blindly isn't the move. But the heavy lifting is done before I even start my morning coffee.

Focusing on What Actually Matters

The biggest impact of all this isn't speed. It's focus. When AI handles the operational noise, I can spend my time on the bigger picture:

• Architecture decisions that affect reliability at scale
• Strategic automation that eliminates entire categories of toil
• Incident prevention rather than just incident response
• Mentoring and collaboration with the team on complex problems

I went from spending a significant portion of my week on tasks that any script or agent could handle, to spending that time on work that actually requires human judgment, creativity, and experience. That's the real value of AI in an SRE role. It doesn't replace you. It clears the path so you can do your best work.

Looking Back at How Fast This Happened

It's honestly surreal how quickly things moved. A few years ago I was quietly using ChatGPT on the side, feeling like I was doing something I shouldn't be. Now I'm building autonomous agents at work with full organizational support and enterprise tooling.

If you're in a position where AI is still discouraged or restricted, I get it. The security and compliance concerns are real and they need to be taken seriously. But if your organization hasn't started building a strategy for AI adoption, they're falling behind. The engineers and teams that embraced AI early are already operating at a different level.

The Takeaway

AI didn't make me a better engineer by giving me answers. It made me a better engineer by giving me time. Time to think, time to design, time to focus on the problems that actually move the needle. The mundane work still gets done. It just doesn't require me to be the one doing it anymore.

If you told me back when ChatGPT launched that I'd be building agents and automating entire workflows with AI in my day job, I probably would have believed you. I just wouldn't have guessed it would happen this fast.