I expected the robots in this book to read as a dated artifact — a 1980s manufacturing plant investing in expensive new machines, told through a lens that wouldn't map to anything happening now.
Instead I found the clearest description I've read anywhere of what happens when an organization celebrates a metric that was never connected to the actual goal.
The plant in the story buys robots to automate part of production. The robots work. Utilization on that resource goes up dramatically, and the reports celebrate it. But the rest of the plant is still constrained by something else entirely, so the robots' brilliant local numbers don't translate into more product shipped. If anything, the robots make the surrounding inventory problem worse, because now that one station produces even faster than the bottleneck downstream can absorb.
AI adoption inside most organizations is on the same track. Knowledge work speeds up — more drafts, more analysis, more code, more communication — and that looks like enormous productivity on an activity report. Whether it's actually relieving the organization's real constraint is a completely separate question that the activity report can't answer.
The shared danger, forty years apart: a technology can make one resource look dramatically more productive while the thing actually limiting the system remains untouched.
Current Hypothesis
Any new capability — robotics in 1984, AI now — will be evaluated first against old local metrics before anyone questions whether those metrics still mean anything.
Not a conclusion — a working idea, revisited as evidence comes in.