The Operations Bridge

Point fourteen. A thought that’s been reorganizing how I think about roles: AI didn’t remove the constraint in knowledge work. It moved it.

Generation got cheap. Evaluation didn’t

For most of my career, the expensive step was making the thing — writing the analysis, drafting the plan, building the first version. Reviewing it was the quick part at the end. AI has quietly inverted that. Producing a draft now costs almost nothing. Deciding whether the draft is right, whether it fits the real situation, whether it’s safe to act on — that’s the part that still takes a person who knows what good looks like.

The constraint didn’t leave — it moved

Anyone who’s worked with the Theory of Constraints knows that relieving one bottleneck just exposes the next one. Speed up a machine that was never the constraint and you don’t ship more — you pile up work in front of the real limit. AI is a massive capacity increase on the generation step. So the queue is now forming in front of judgment: the ability to look at a plausible-sounding output and say “no, that’s subtly wrong, here’s why.”

Develop for discernment, not just usage

Most AI training I see teaches the generation step — how to prompt, which tool, what buttons. Almost none of it teaches the evaluation step, which is now the scarce one. The people who will be most valuable over the next few years aren’t the fastest prompters. They’re the ones with enough domain judgment to catch a confident answer that happens to be wrong, and enough backbone to say so before it ships.

A question for leaders

Look at how you’re developing your people around AI. Are you only teaching them to produce faster, or are you also teaching them to judge better? The first is easy to buy. The second is the actual constraint, and it’s the one worth investing in.