James C. Scott opens Seeing Like a State with a forest. In late eighteenth-century Prussia and Saxony, foresters wanted to predict how much timber the state would earn each year. A natural forest was too messy to count. So they replaced it with straight rows of one species, mostly Norway spruce, planted and cut on a schedule.
The first rotation went beautifully. The second generation of trees did badly, and the Germans later gave the decline a name, Waldsterben, forest death. The undergrowth, the dead wood, the birds and the insects had been doing work nobody measured.
Legibility has a price
Scott calls this legibility: rearranging a complicated reality so a central office can read it. He follows it from permanent surnames to planned cities like Brasília.
Every enterprise AI programme runs on legibility. An agent can act only on what someone wrote down and structured. Work that lives in hallway knowledge and favours between colleagues never appears, so a redesign built on the documented process can strip it out the way the foresters cleared the undergrowth.
Scott borrows a Greek word for that missing knowledge, mētis, the practical skill people learn by doing. Before automating a process, ask the people who run it what they do that the procedure never mentions.