I Tried Claude and Copilot in Visual Studio Code at My Corporate Job. Here's What Happened

In my private projects I always reach for Claude. At my corporate job I started out with GitHub Copilot inside Visual Studio Code — not because I'd compared anything, but because I assumed it was the only AI I was allowed to use. Then I found out Claude was available there too, gave it the exact same task, and the difference was hard to ignore: on one PL/SQL procedure, Claude did with about a third of the effort what Copilot had taken a long, winding road to reach.

At Work I Reached for Copilot — Because I Thought It Was the Only Option

Outside of work, Claude is my default for everything, coding included. So when I sat down at my corporate job, my starting point wasn't a considered choice at all: I opened Visual Studio Code, reached for GitHub Copilot, and got going — genuinely believing it was the one AI assistant I was cleared to use. That assumption is the whole reason this comparison happened later than it should have.

One PL/SQL Procedure, a Lot of Tokens, and Ugly Code

The task itself was specific and real: I had a PL/SQL procedure that needed a few changes, and I set out to refine it with Copilot. It worked — eventually. But getting there cost me a lot: I burned through a huge number of tokens, went around in circles more times than I'd like to admit, and what I ended up with was, frankly, ugly and not very optimal. That last part isn't just my opinion — it's what the PL/SQL expert who looked at the code told me too.

Here's the nuance I want to be honest about. For a side project, I'd have shipped it without a second thought — it did what I needed, and that's often enough. But corporate work is a different bar. The code has to be much cleaner and tighter, and by that standard I just wasn't happy with what came out.

Then I Remembered Claude Was Right There

The turning point was almost embarrassingly simple: I realised my beloved Claude was available to me at work as well. So, without overthinking it, I handed Claude the exact same task. It simplified the code dramatically, got it doing precisely what I wanted — and it did all of that with roughly a third of the effort Copilot had demanded from me. Same goal, same working result, a fraction of the back-and-forth.

Was It Just Better Prompts? I Don't Think So

The one doubt I keep circling back to is fairness: by the time I asked Claude, I understood the procedure better than when I first started with Copilot, so maybe my prompts were simply sharper the second time. It's a real question, and I don't want to pretend it away. But no — I genuinely don't think that's what happened. My honest read is that Claude is just much better at code than Copilot.

The Bottom Line

I started with Copilot at work only because I assumed I had to, and it did get the PL/SQL procedure done — but at the cost of a lot of tokens, a lot of circling, and code that wasn't good enough for a corporate standard. Claude, handed the identical task, gave me cleaner code and the result I wanted with about a third of the effort. I can't fully rule out that knowing the procedure better helped my prompts the second time, but I don't believe that's the explanation. For me, on this task, Claude was simply the better coder.

FAQ

Claude or Copilot for coding — which was better?

In my experience, on the same PL/SQL procedure inside VS Code, Claude was clearly better: it produced far simpler code and reached the result I wanted with about a third of the effort GitHub Copilot needed from me. I think Claude is simply much better at code.

Is GitHub Copilot good enough for corporate work?

Copilot did get the job done — but it burned a lot of tokens, went around in circles, and the code was ugly and not very optimal, according to the PL/SQL expert who reviewed it. For a side project that would be perfectly fine; for corporate work, where the code has to be much cleaner, I wasn't happy with it.

Did Claude only win because you knew the procedure better the second time?

That's the one doubt I have, since my prompts may have been sharper by then. But I don't think it's the real reason — my honest conclusion is that Claude is genuinely better at code than Copilot.