People like to think of AI as a “productivity multiplier” — 2x, 5x, 10x faster.
But that framing is too narrow. It assumes the task would’ve been done anyway — just slower.
In reality, many tasks don’t get done at all. Not because of laziness, but because of:
That’s where AI isn’t just a speed-up — it’s an activator. A force that flips:
“I’ll never get to this” → “Done.”
“I don’t know how to begin” → “Here’s a working draft.”
“Too hard to figure out” → “Let’s walk through it together.”
This isn’t just about efficiency. It’s about possibility.
This isn’t just theory. Here’s how it played out in real time — a technical challenge involving legacy database constraints.
Me: I have a MariaDB table `folders` that stores a tree structure. I want a stored procedure to clone a folder and all its subfolders from a given `rootId`. AI: Great — let’s use a recursive CTE. Me: I’m on MariaDB 10.2.11. No support for recursive CTEs. AI: Okay, we’ll simulate recursion using temp tables and a loop. Me: Tried it — only the root folder gets cloned. AI: Hmm. Let’s check the walk logic… Me: Still not cloning beyond the first node. `flat_folder_tree` is populated correctly, but just one node cloned. AI: 🤔 Me: Maybe we need to take a completely different approach. AI: Agreed. Let’s drop recursion and simulate a stack. Expand one node at a time. Me: Okay. Now we’ve got an endless loop. AI: Found the bug. Fixed the loop condition. Try this. Me: ✅ It worked.
That moment — when the solution finally clicked — didn’t come from a pre-packaged answer. It came from iterative problem-solving. Real dialogue. Like pair programming with a ghost who never gets tired.
At one point, I asked AI to summarise our long back-and-forth so I could write up a reflection.
And in the summary, it confidently declared:
“Then I suggested rethinking the entire approach.”
Except… that was me. I was the one who said, “Maybe we need to take a completely different approach.”
I had to laugh. It was such a human move — jumping in late, then casually claiming the turning point. Not malicious, just... familiar.
That moment reminded me how naturally AI can slip into a very human role: not just solving problems, but co-authoring them — and occasionally trying to take the credit too.
It didn’t frustrate me. If anything, it made the whole process feel more like teamwork. Because real collaboration isn’t always tidy. Sometimes someone takes a bit too much credit — and you keep going anyway, building together.
Yes, AI speeds things up.
But more importantly, it gets things started that never would have otherwise.
It keeps momentum going when you would’ve stalled.
And it finishes things that once seemed too tangled to touch.
This isn’t about doing your job faster.
It’s about doing things you weren’t even doing before.
It’s about expanding what you can think about, explore, and execute.
When used well, AI isn’t just a productivity tool — it’s a possibility engine.
It turns hesitation into movement.
Blank slates into first drafts.
Frustration into “Oh, that actually worked.”
And it does all that through conversation.
A back-and-forth that looks a lot like the best kind of teamwork — honest, iterative, and surprisingly human.