During my internship in Perth, I quietly rebuilt a 3-year ERP project in 3 days. Not by working harder — by deploying a workflow I’d already built.
I finished my internship a couple of months ago. It wasn’t mandatory — optional, actually — but I took the chance. I was placed as a developer at a company here in Perth, and on day one I was handed a task: continue building an ERP system.
You know the kind. One dashboard where a business can manage projects, clients, documents, budgets — instead of juggling five different spreadsheets and hoping nothing falls through. Classic software project. The previous intern batch had already kicked it off, so there was a design, there was a direction. Our job was to continue and finish it.
The timeline on the project: three years.
The team were experienced developers. But the way they were working — the process — was holding everything back. I could see it clearly within the first week. At the pace things were moving, three years wasn’t a guarantee. It might’ve been a starting point.
I had an idea. What if I used my own AI workflow and rebuilt the thing properly?
I know how that sounds. And I wasn’t going to blow everything up overnight — that would’ve shocked everyone: the team, the boss, the whole project. That’s not how you earn trust. So instead, I kept doing the project the way we’d all agreed on. But on the side, quietly, I was recreating the whole application using my own workflow.
I finished a working prototype in three days.
Three years became three days.
When the boss saw what I’d built, he changed the direction of the whole project. Some things on the original list — we crossed them out. And because things were moving so fast, we just kept building on the prototype. It ended up being what everybody was working toward. I finished my internship with solid results, and I got offered to stay on after I graduate.
Here’s where I have to be straight, because this isn’t a story about AI being magic.
The prototype was good for two specific reasons that had nothing to do with the tool itself.
First: I wasn’t guessing from zero. The previous intern batch had already done the design work — I had wireframes, a direction, a visual language to work from. That’s not a small thing. When you’re prompting an AI to help you build something complex, having a clear design to reference is the difference between getting something coherent and getting noise.
Second: I’d already built similar projects before. I knew roughly what the architecture should look like for this kind of application. I wasn’t working out the fundamentals as I went — I already had them.
So the prototype was fast because I had experience. AI multiplied that experience. It didn’t replace it.
My fellow interns were using AI too. Same tools, roughly. But they were using it the standard way — the safe way, you might say. Staying within what they already knew AI could do. Safe prompts, safe outputs, safe results.
I’d spent time before the internship actually working out what AI is capable of doing for this kind of project — and I’d baked that into my workflow. Every step of the process already had AI in it. I didn’t have to spend weeks figuring it out during the internship, because I’d done that work before I even got there. So when the moment came, I just deployed what I’d already built.
That’s the gap. And it’s bigger than it looks.
There’s a difference between using AI to generate code and using AI to actually solve a problem.
A good solution depends on so many variables — the problem itself, the environment, the budget, what the client actually wants, what the team can maintain. Code is just one part of that. A pretty small part, actually.
And that’s where a workflow matters. Because to solve the problem, you have to consider all of those variables, not just generate code. What I’d built across all those different chapters of my life — the business thinking, the marketing instincts, the years of shipping — got me to the actual solution. Not just the code for it.
Software engineering is problem solving. Not just coding. Most people, developers included, are working with AI as they go. They have the tool. They don’t have the workflow.
Don’t wait for a big project to start figuring out your AI process. Build it now. Refine it. Make it yours. So when the moment comes, you’re not working it out on the spot — you’re deploying what you’ve already built.
Three years became three days. Not because of a tool. Because of everything that came before it.