Once I started looking for it, I found queues everywhere that nobody in the building called a queue. Approvals sitting in an inbox. Analyses waiting for a meeting that kept getting pushed a week at a time. A recommendation that had been technically "done" for a month, just waiting for someone with the authority to say yes.
None of it looked like inventory, because none of it was sitting on a shelf. It was sitting in someone's calendar, or someone's inbox, or the gap between two people who each assumed the other had it.
Calling it decision inventory instead of "just how things are" changed how urgently I wanted to move it. Inventory has a cost. It sits there depreciating, tying up capacity, blocking whatever's supposed to happen after it. Decisions waiting to be made do the same thing, just less visibly, because nobody's counting them.
The uncomfortable question this raises for AI adoption specifically: if AI keeps making the knowledge-production side faster, and the decision-making side doesn't get any faster, the queue doesn't shrink. It grows, just further along in the process, where it's harder to see and easier to mistake for progress in the meantime.
This is still an open question for me — I don't yet have a clean way to measure decision latency the way a factory measures cycle time, and I'm suspicious of the hypothesis until I do. It's in the Research Lab, not settled.
Current Hypothesis
Decision latency, not knowledge production, is the more common constraint in organizations that have already adopted AI tools.
Not a conclusion — a working idea, revisited as evidence comes in.