I try to avoid justifying my own existence, but its a reason why I think skilled engineers are way more productive using LLM's than person off the street. Because they can confine them. Though there is no technical reason that ability can't be codified.
"Slower" is doing a lot of work there. It's the difference between an engineer or two wrecking havoc doing a major refactor and convincing a manager that it's ok while you're away on vacation, and the same engineers failing to convince the manager because it will take forever, and quietly dropping the project after hitting the first roadblock.
"How long is this going to take?": "a week" vs "a quarter" makes a huge difference.
Extreme speed is a form of quality; the fear is that it enables bad things to happen.
I've been solo prenuering for the last year so didn't have a lot of data from others but your comment convinced me its a problem worth solving, so I put this together https://vitruvius.8by3.com/
Its hard wip, pushed seconds ago, paint still wet. But the idea is to "enforce" good architectural practices for your agents. Can store constrains and ADRs in git so everyone in a team (everyones agents) can read from the same page.
AI still needs lots of guidance, and might seem like a genius that does everything on its own, until you give it to people who can't guide it and things begin spiraling out of control.