01
The 40% Question Is Wrong
A business-wide opinion about AI relevance is too large to guide a useful decision. Start with memory loss, repetitive information movement, and warnings that arrived too late.
Practical notes for people turning AI into real work.
The website is the durable index. These notes begin with the work itself: where information is lost, where review becomes theatre, and whether a workflow improves after the first impressive run.
Questions that make the hidden work visible.
02
The Second Run Is the Real Test
The first output is not proof of value or failure. Track correction, exception, review, and downstream work to see whether an AI-assisted workflow is learning or merely moving effort.
03
Human in the Loop Is a Job, Not a Slogan
A person is not a control just because they receive an output. Real oversight starts with a named owner, visible evidence, and authority proportionate to the consequence.