What AI Can See — and What It Cannot

Machine intelligence works on what has been captured: text, numbers, transactions, documents, signals that someone chose to record. Within that boundary it is faster and more consistent than any human analyst.

The boundary itself is the issue. Most of what determines whether a consequential decision holds is not in the record: what a partner intends but has not said, what a team has quietly stopped believing, what an environment has absorbed from everything that happened in it before.

This is not a criticism of the tools. It is a description of their input. A model cannot weigh what was never written down.

So the question in a serious situation is not whether the analysis is good. It usually is. The question is what sits outside it, and whether anyone in the process is responsible for looking there.