"The AI Can Do It" Is Not an Excuse To Tolerate a Mess

As a follow-up to yesterday's post, a reader wrote in with a concern: That organizations might use AI as an excuse to accept sub-par processes. The example: What if a particular task requires cross-referencing ten different spreadsheets? If you let a human handle that task, they'll eventually get fed up and do some consolidation work. An AI, however, will not complain and will just happily burn compute and ramp up usage on this inefficient task.

And as time progresses and the process gets more and more complex, with agent upon agent heaped on it, it eventually becomes utterly unmanageable by both humans and the AI, creating an unhealthy dependence on the AI, where it's now so complex that you have to throw the whole thing away and start from scratch.

Good software engineering teaches the way out of this: Whenever you interact with a system, look for little ways to leave it better than you found it. These small, opportunistic improvements (which are now even easier thanks to AI) go a long way in keeping everyone sane.