How I'm using AI as a Cloud Engineer

Almost daily, I’m asked about how I use AI in my work and everyday life.

It has been over 3 years since the ChatGPT moment happened, and it feels as though we are all still trying to understand how AI fits into our everyday lives.

I don’t have the answer, but I can share how I approach it as a Cloud Engineer working in enterprise systems.

From my point-of-view, the number one way to use AI right now isn’t necessarily to write or generate code, but to learn and understand concepts.

In this blog, I lay out why this is my preferred way of using AI in the age of vibe coding and AI-enabled development.

I think that by now, most people have come across the term vibe coding.

If you haven’t, I’m not sure where you were last year as this was one of the most popular buzzwords that frequented the tech scene in 2025.

The premise of vibe coding is that you have AI generate all the code, whilst you as the engineer are behind the wheel coordinating the whole thing via natural language.

If you’ve written code prior to 2022, I don’t need to convince you of just how fascinating this advancement is.

The issue with vibe coding, however, is that it is only good for trivial use cases or low-stake situations, e.g. writing scripts or building a PoC/MVP.

Essentially, cases where failure is acceptable and can be easily managed.

In an enterprise environment, I believe a greater sense of duty of care is required.

The systems that we work on affect the lives of individual people: their personal data, money and the trust they have placed in the organisation.

As a cloud engineer, SRE or any engineer for that matter, our job isn’t simply to accept whether something works but to account for the worst-case scenario, including the impact of such failure.

Keeping this in mind, vibe coding or AI-enabled coding doesn’t quite fit the bill.

But there is a middle ground...

So, here is how I use AI in my daily job and what I think is the most appropriate way to leverage AI in its current form.

First and foremost, I always disable any feature that enables AI to change the codebase on my behalf, i.e. Agent mode. I don’t want AI to make any changes that I haven’t thoroughly reviewed or understood myself.

I use only ever use Ask or Chat mode, and restrict my usage to two cases:

1. Explaining concepts I want to further understand  
2. Generating code that I review and manually integrate

Using AI in this way gives me confidence as an engineer that I know exactly what changes are happening before they have the potential to affect anyone else.

However, there is a caveat to this.

If I know the change I am making has relatively low/minor impact, then I might lean on AI more heavily, e.g. writing a bash script that makes some part of my job easier.

There is no hard and fast rule for this, and a big part of what we do as engineers is knowing which tools to use and which not to use at any given time.

AI is no different.