I've lived on both sides of the ops equation.
Early on, I ran other people's systems. You learn fast that processes look great on paper. The problem is paper cuts. The edges are where you bleed. The handoffs, the exceptions, the "this only happens on Tuesdays" problems. That's where the real education is.
When I started building the systems myself the challenge became choosing what actually matters. Deciding what gets measured, what gets automated, and just as critical, what gets left alone.
Different vantage point.
But here's what I didn't expect about building in the AI era: the friction is so dramatically reduced on everything; communicating, writing software, doing deep analysis, just thinking alone, that you start to feel like speed is always the answer.
So you try to do everything. At once.
At this stage I think we've all caught ourselves building things that looked great in the first week and got muddy by week three. Ambitious projects that required rethinking everything once the easy part was done. Half-baked tools that solved problems that are already solved.
Meanwhile, some of the most impactful things I've shipped are embarrassingly simple. A QR code generator. A pricing label tool. An automation that saved someone forty-five minutes a week. Clear problem, obvious solution, done.
This has become a bit of a mantra for me: I can do anything. But I can't do everything, and I can't do it all at once.
I think the counterintuitive move with AI, especially right now, when everything feels possible, is to intentionally slow down sometimes. Scale back. Some of this is so new that it's going to take time to understand how to even think about it properly.
Speed is the right answer for solved problems.
For everything else, the friction is doing some useful work.