Both pieces come out of the same build. One is about what a system has to do when it cannot be sure, and one is about the order the work has to happen in. They are written for people who have to defend the decision afterwards.
Writing
Positions, argued from what shipped.
-
What does it do when it cannot be sure?
Four rules for building AI that has to be provably right: every claim shows its source, the model chooses instead of writing, boring mechanisms where they can be checked, and refusing to publish with a caveat.
-
Enterprise AI is an order-of-operations problem
AI literacy, a workflow that touches revenue or cost, a feasibility study, then a task list where the who is derived from risk rather than chosen. Adoption is a sequence, not a licence purchase.