Definition
The structured record of how one specific business actually works: its services and real prices, its customers and their history, its policies, its exceptions, the promises it has made, and what should happen next. It is what lets an AI system answer a routine question correctly, because the answer comes from that business's own records rather than from a general model.
What it looks like in practice
A pet grooming salon gets a text on a Sunday: 'Can you do Bailey next Saturday, same as last time?' Answering needs four facts nobody has written down anywhere useful, which dog Bailey is, what 'same as last time' was, which groomer handles him because he is anxious with clippers, and how long that groom actually takes. With those recorded, the reply is accurate in a minute and the slot is held. Without them, the message waits until Monday for the one person who knows, and the pet parent has already texted somewhere else.
Why it matters
This is the quiet difference between a system that helps and one that generates confident nonsense. A general model knows what grooming appointments are like; only a business's own record knows that this salon does not book aggressive dogs after 3pm. Two things keep it honest: the record has to be correctable by the owner in plain language, and the system has to say when memory does not cover a question instead of filling the gap. Memory that cannot be inspected is just a more expensive guess.