I ran into this with what should have been a simple task.
Once a month, I take a few Excel files, transform them, and load them into an accounting system. The process is straightforward and predictable, making it an ideal candidate for AI assistance.
So, I leaned into it, and the results came quickly. Instead of starting with a blank file, I had working code almost immediately. The inputs, outputs, and transformations were already in place.
Then it started adding things I never asked for.
It introduced an orchestration layer, added checkpoints between steps, and built in restart logic and fault tolerance as if this script were part of a much larger, always-on system. Individually, none of those ideas were bad. In the right context, they’re solid patterns.
But this wasn’t that context.
This script runs once a month. If it fails, I rerun it. The entire process takes a few minutes. All of that additional structure provided little value. Instead, it made the solution larger, more complex, and more expensive to maintain.
That’s when something clicked.
AI didn’t eliminate the usual tradeoffs. It amplified them.
We’ve always operated within the same constraint: time, cost, and quality exist in tension. Move faster and something gives. Cut costs and risk tends to show up somewhere else. Push for higher quality and it usually requires more effort.
AI doesn’t change that reality. It simply accelerates everything.
Including mistakes.
AI is powerful, but it isn’t always right. And when something that is “mostly right” operates at high speed, small issues rarely stay small. Patterns get reused, assumptions get copied, and decisions get embedded in more places before anyone stops to question them.
In traditional workflows, there’s natural friction. Teams pause, review, and catch problems early. AI removes much of that friction. As a result, errors don’t just happen faster, but they also scale faster.
It’s like following a GPS that recalculates instantly and keeps pushing you forward. It feels efficient, smooth, and mostly right. But if the route is off, maybe it missed a road closure or suggested a shortcut that doesn’t actually work, you don’t just make a single wrong turn. You keep going in the wrong direction longer, because the system is confidently guiding you there.
The best way I’ve found to think about AI is simple: it’s a very fast junior resource.
It can do a lot, it moves quickly, and it’s often impressive. But you still review the work. You still make sure it understood the problem correctly and chose the right approach.
AI behaves the same way.
It tends to apply patterns it has seen before, whether they actually fit or not. The difference is volume. A junior developer might produce few meaningful pieces of work in a day. AI will generate dozens. When it’s moving in the right direction, that creates tremendous leverage. When it’s not, you’ve generated a significant amount of rework in a very short period of time.
That’s the real shift: creating the first draft is cheap. Evaluating it is where the value is.
You end up spending more time asking questions like:
- Does this actually solve the problem?
- Is this the simplest approach, or just the most elaborate one?
- Is this something I want to maintain six months from now?
Those questions matter more than ever because execution is now cheap. Judgment isn’t.
AI can absolutely help you move faster. It can reduce effort, and in the right environment it can even improve quality. But it doesn’t eliminate tradeoffs. If anything, it forces you to confront them sooner.
If you push for speed, problems surface faster. If you optimize for cost upfront, you may pay for it later in rework. And if you care about quality, you still need oversight and review. That hasn’t changed.
That monthly script was a good reminder. AI got me moving quickly, which was valuable. But it also nudged me toward overengineering something that didn’t need it.
If I had followed it blindly, I would have spent more time and money maintaining something that was supposed to be simple.
AI isn’t an expert system. It’s a multiplier.
And multipliers make both good decisions and bad decisions bigger.
So no, AI doesn’t break the time–cost–quality triangle. It amplifies it. And the faster it helps us move, the more intentional we need to be about the direction we’re moving in.