AI has made polished conclusions cheap. A confident memo is no longer evidence that anyone understands the problem.
Alchemy became chemistry through a change in protocol, not a leap in intelligence. Alchemists hid methods in metaphor, so discoveries often died with them. The Royal Society imposed a harder standard: publish the method, let others reproduce it, and allow public correction.
Most organizations still practice epistemic alchemy. They circulate the answer while hiding the inputs, assumptions, tests, rejected alternatives, and revision history. AI makes this failure harder to detect because the output can look finished before the thinking has started.
The real unit of transparency is not the conclusion. It is the reproducible, patchable method behind it.
If another person cannot rerun the work, locate the error, and improve the method, you do not have organizational knowledge. You have an artifact.